Representation of Product Hazards in Consumer Memory

Arno J. Rethans, The Pennsylvania State University
Manoj Hastak, The Pennsylvania State University
ABSTRACT - Twenty consumer subjects were probed for their hazard related knowledge structures. A sequential free elicitation procedure attempted to differentiate between linguistic thoughts and image-based thoughts. These thoughts were next classified as episodic or semantic so as to provide some indication of the content of the knowledge structures. Findings suggest that research on information representation will almost always be incomplete if it focuses on linguistic information only and that the research should not exclusively focus on representation in semantic memory. Recommendations for future research are provided.
[ to cite ]:
Arno J. Rethans and Manoj Hastak (1982) ,"Representation of Product Hazards in Consumer Memory", in NA - Advances in Consumer Research Volume 09, eds. Andrew Mitchell, Ann Abor, MI : Association for Consumer Research, Pages: 487-493.

Advances in Consumer Research Volume 9, 1982      Pages 487-493


Arno J. Rethans, The Pennsylvania State University

Manoj Hastak, The Pennsylvania State University


Twenty consumer subjects were probed for their hazard related knowledge structures. A sequential free elicitation procedure attempted to differentiate between linguistic thoughts and image-based thoughts. These thoughts were next classified as episodic or semantic so as to provide some indication of the content of the knowledge structures. Findings suggest that research on information representation will almost always be incomplete if it focuses on linguistic information only and that the research should not exclusively focus on representation in semantic memory. Recommendations for future research are provided.


Researchers in the area of hazard assessment, product safety and public policy formulation have recognized that any hazard management system developed to protect consumers from the risks associated with products and/or technologies must take into consideration both the technical and the psychological/social facets of risk assessment (Slovic et al. 1981, Slovic et al. 1977; Otway 1973; Rethans and Albaum 1980; Rowe 1977). The technical facet involves the identification of hazards and the measurement of their probability and consequences; whereas the psychological/social facet involves the public perceptions of the risks and the acceptability of these perceived risks.

The recognition of the psychological facets of hazard management is crucial to the success of public policy programs (Thomas et al. 1980). As Slovic, Fischhoff and Lichtenstein (1979) warn; "People respond to hazards they perceive. If their perceptions are faulty, efforts at public and environmental protection which to not consider these perceptions are likely to be misdirected." In the area of consumer product hazard management, Rethans (1980) and Rethans and Albaum (1981) have argued similarly for the introduction of consumer hazard perceptions into both private (i.e. company) and public product hazard management programs.

Studies of the validity of risk perceptions have obtained some mixed results. Indeed, the empirical work has culminated in the identification of some interesting "lay foibles and expert fables" in judgments about risks (Fischhoff et al. 1980). In some cases, people are quite attuned to risk information; in other cases, however, their judgments show strong and systematic biases (Lichtenstein et al. 1978; Fischhoff et al. 1979; Combs and Slovic 1979; Rethans 1980). Of the variety of explanations offered for the observed biases, the more intriguing one is the notion of the availability heuristic (Tversky and Kahneman 1973). When employing this heuristic risk judgments are based on information that is easily activatable. Biased judgments under this perspective, then, are a function of biased information representation and/or activation.

From an information processing perspective these empirical results stimulate some interesting questions about public or consumers' hazard related cognitive structures and cognitive processes. What is the content and structure hazard knowledge representations in memory? What factors influence the activation and further processing of stored information? Could understanding of hazard information representation and activation lead to a better description of the cognitive processes underlying availability and other observed heuristics? Answers to these and similar questions will have both theoretical and practical implications for the study of judgmental processes in general and for research in risk assessment in Particular.

The study reported here represents an initial attempt to characterize the content of consumer representation of product and technological hazard information in memory.


Knowledge Structures

The theoretical framework underlying the study presented here is based on the associative network representation of knowledge (Anderson and Bower 1973) and the spreading activation theory of processing (Collins and Loftus 1975).

In the associative network model, individual concepts are represented as nodes in a network and collectively shape the content of the knowledge structure. On the other hand, the links between the nodes represent the associations between the concepts and account for the organization of the knowledge structure. Activation of the resulting network is initiated by an internal or external cue starting at a single concept node and spreading along the arcs linking the activated concept to other nodes at a decreasing gradient. This decrease in gradient is hypothesized to be inversely proportional to the accessibility or strength of the links in the path. Thus, the theory suggests that concepts that are strongly linked to the initially activated concepts are likely to be activated, while the decreasing gradient allows the activation to "die out" over time without activating more than a small subgraph of the entire associative network (Ortony 1978).

The distinction between content and organization has been introduced as a vehicle to characterize existing research on consumer's knowledge structures. Johnson and Russo .(1978) illustrate research emphasizing the organizational aspects of knowledge structures. Chronometric analysis as well as analysis of recall order led these authors to speculate about the organization or product information in memory. In contrast, the 1980 Russo and Johnson study reflects interest in both contentual and structural aspects. The coding scheme developed to classify consumers' knowledge about common products incorporates one content-based variable (the inferential levels of knowledge) and one structure based variable (brand-based versus attributed based organization). Finally, Kanwar, Olson and Sims (1981) propose and provide empirical indicators for three content-based measures of knowledge dimensionality, abstraction and articulation - which provide more generalized, more abstract descriptions of memory content. Interest in content measures of knowledge structures is based on the assumption that they may provide additional, if not better, explanations of subsequent consumer information processing activities (Brucks and Mitchell 1981). In the study reported here we continue the interest in the content of knowledge structures by examining the content of consumer hazard knowledge structures.

Content Typology

The content typology to be employed in the classification of activated hazard-related knowledge is based on the findings of research in the area of risk assessment. Specifically, research findings related to the role of experiential knowledge and findings related to the role of image based hazard representations in the formation of risk judgments led to the selection of the content typology. Before describing the content typology, we first briefly outline the relevant research findings.

Experiential versus Inferential Knowledge. Researchers in the area of risk assessment have suggested the importance of individual experiences in the formation of risk judgments. For example, Rates (1962) observed that residents of flood plains appeared to be "prisoners of their experience," unable to conceptualize floods that never occurred or to see the future as anything but a mirror of the recent past. Lichtenstein et. al (1978) show a strong correlation between extent of experience (both direct and indirect) with death and injury events and the subjective estimates of levels of risk associated with these events. Finally, Abelson in reviewing Slovic, Fischhoff, and Lichtenstein (1976) seems to suggest that episodic script processing may be responsible for the observed distortion of probability judgments. These findings and speculations imply that hazard information is represented as encoded records of experienced episodes rather than generalized inferences derived from such episodes.

In an attempt to capture the experiential versus inferential basis of knowledge representation in our content typology, we turned to the distinction between episodic and semantic memory introduced by Tulving (1972). Episodic memory receives and stores information about temporally dated episodes or events, and the temporal-spatial relations among these events. Episodic memory is a more or less faithful record of a person's experience. Examples of episodic hazard memory would be the following statements made by two of our subjects. "I remember, in third grade, I was riding my bike with no shoes on and my foot caught in the spokes and I cut my toe really bad." "I had a hair dryer some time ago that shocked me every time I used it, so my husband had to tear it apart to see what the cause was." Each of these statements refers to a personal experience; they are autobiographic in nature. Semantic memory on the other hand contains one's general abstracted knowledge about facts and principles. It preserves people's general conceptual information, world knowledge, and linguistic abilities (Lachman, Lachman and Butterfield 1979). Illustration of hazard information contained in semantic memory include: (a) ammonia should not be mixed with clorox, it will create deathly fumes, (b) some of the hair dryers on the market contain asbestos, and (c) children should be kept in the house when you mow the lawn because the lawnmower will kick out rocks and twigs that will hurt the kits.

The semantic-episodic distinction is employed here as a conceptual tool rather than as a functional distinction in human memory. Although there is an interesting controversy regarding the psychological reality of these two memories (Anderson and Ross 1980; Hermann and Harwood 1980), of greater value to this paper is the realization that semantic and episodic memories are not always cleanly separable. They may be considered instead as end points on a continuum ranging from completely context-dependent episodes to truly general knowledge (Olson 1980; Kintsch 1980). This perspective does not impair the usefulness of the distinctionS however, as will be shown later.

Linguistic versus Imagery Based Representation. The idea that people code events both linguistically and perceptually seems to be accepted by most psychologists (Paivio 1971; Posner 1973; Rossiter 1976; Shepard 1978; Wickelgren 1981). Indeed, mental images and their vividness are perceived to affect subjective risk assessments. For example, Slovic, Fischhoff and Lichtenstein (1981) find nuclear power hazard perceptions to be influenced by vivid mental images of a nuclear accident. Similarly, in the area of consumer behavior. Rossiter (1980) has argued that episodic images of past consumption or usage situations may form important components of product memory that may be activated in choice situation. Hence, for purpose of this study we further classify memory content into linguistic versus visual representations and focus once more on the content of these representations, i.e., the content of the reported hazard images.


Study Procedures

Subjects were run individually through a single study session which was comprised of two segments. In the first segment of the study a free elicitation procedure (Olson and Muderrisoglu 1979) was employed to identify the hazard knowledge that consumers have stored in memory. Subjects were directed to verbalize all of the hazard-related thoughts, sensations, feelings and/or images that were generated in response to several hazard cues. Their verbal protocols were unobtrusively tape recorded for later analysis. Upon completion of this segment, the respondent completed a short questionnaire containing a 7-point scale designed to measure the extent to which pictures or images had entered their thoughts, as well as a series of scales evaluating the vividness of the evoked imagery for each of the stimuli (Marks 1973). The latter set of scales have previously been found to be both reliable and valid (White, Sheehan and Ashton 1977).

In the second segment, the respondent was asked to elaborate on the pictures ant/or mental images reported in the early part of the interview. Respondents were instructed to refer back to the images they had experienced and describe them in more detail. Verbalizations were again tape recorded for later analysis. No respondent reported any difficulty in executing this task.

This image elicitation procedure does not completely rule out the possibility that respondents constructed images during the second verbalizing task rather than merely elaborating upon images actually experienced during the first part of the study. Several precautions were taken to minimize the effect of this possible demand characteristic. First, protocols obtained during the second segment of the study were matched with those obtained during the first segment so as to ensure that subjects were indeed elaborating upon previously reported images. Secondly, as illustrated later in Table 2, a very conservative coding scheme was used to identify image based responses present in the protocols obtained in the second segment. Only responses containing such clear phrases as "I saw..." or "I had this picture..." were coded as image responses. Lastly, the combined results of the two sets of protocols and the use of multiple scales to measure the extent of mental imagery elicited provide some degree of confidence in the results to be reported.

Finally, the subject completed a questionnaire which incorporated the 15-item Visualizer-Verbalizer Questionnaire (W Q) developed by Richardson (1977), perceived personal and societal hazardousness scales, indicators of use and injury behavior data as well as the more standard demographic measures. Included also were items designed to elicit subjects' reaction to the tasks. Most subjects indicated that they found the tasks to be relatively easy, just right in terms of length, and natural rather than artificial. Furthermore, the participants felt that the information they provided was both accurate and representative.

Protocol Coding

The free elicitation protocols for both parts of the study were first divided into short segments representing a complete thought or a complete image respectively. Both authors served as independent judges for this part of the coding process in which 1044 thoughts and 307 images were identified. The two coders initially agreed on 875 of the thoughts and 285 of the images.

Elements for the coding scheme to be employed in the classification of the thoughts and the images were developed on the basis of the theoretical frameworks discussed earlier and by scanning part of the actual protocol data. The resulting coding scheme for the thought protocols is shown in Table 1, which the coding scheme for the image protocols is shown in Table 2.





The basic structure of the hazard thoughts coding scheme is derived from Tulving's (1972) notion of episodic and semantic memory, but interpreted as a continuum as suggested by Kintsch (1980). The episodic endpoint of the continuum suggests knowledge stored in memory that can best be characterized as the encoded representation of the consumer's personal experience of a hazardous event often involving negative consequences; we have labeled this segment personal episodic. Autobiographical references can also be made to events in which the negative consequences accrued to significant others; these have been labeled episodic. The semantic knowledge, i.e., the general abstracted knowledge about hazards associated with products, was also classified in two categories. The personal semantic category involves semantic knowledge to personal and significant others' behavior. The semantic category the more abstract representation of hazards reported by the subject; the more general knowledge. The final category, the affect category, initially came out the protocol scan and is of theoretical interest as shown by the works of Zajonc (1980) and Bower (1981). Affect expressions often occurred as an initial reaction to the product cues. (The same five categories were employed to code the non-hazard related events.)

The imagery protocol coding scheme, shown in Table 2 indicates that hazard related images were first classified as involving a single image or apparently involving a chain of vignettes. Within each of these classes, four subcategories were developed based on a scanning of all the images reported. The first two subcategories predominantly contain the imagery associated with the personal episodic and episodic knowledge categories. Images were also reported without the personal reference of "my," such as my son Todd, my neighbor, my Christmas tree, etc. These images were coded as neutral; if the image was reported as having been induced by newspaper or television reporting or a commercial, the image was coded as neutral-media. This last category was created to investigate the extent to which vivid visual portrayal of accidents by the media has an impact on representation of hazard knowledge.

Hazard Cues

The product stimuli used as probe cues in the elicitations were to be highly familiar so that a typical consumer would be able to verbalize the knowledge structure associated with the probe concept. In addition, the products were to represent the range of frequency and severity of actual product hazards as determined from the data maintained by the Consumer Product Safety Commission. The products chosen on the basis of these criteria were power lawn mower, hair dryers, knives, bicycle, Christmas lights and household ammonia. The familiarity ratings on these products provided by the respondents on the typical 7-point scale ranged from 4.52 (ammonia) to 5.81 (knives). Practically all respondents indicated that they had used the products.

In addition to these product stimuli, two technologies were also used as cues, namely commercial aviation and atomic power. These two technologies have been included in many of the hazard and risk assessment studies referred to in the introduction. They were included to enable interesting comparative analyses.


The participating respondents constitute a convenience sample of 20 women staff members of a major Eastern university. Sixteen were married, while three were single; 17 had children. Their educational experience was quite varied. Of the subjects that had finished college and/or had done post graduate work, eight of them had finished high school and the remaining seven completed some college. Total household income showed a distribution ranging from less than $7500 to over $30000, with the modal class being $20000-$30000. Each participant was paid $5 for completing the one-hour session.


Thoughts and Images

The sample total number of thoughts as well as the average number of thoughts per subject generated in the first part of the study are reported in Table 3. For the six consumer products a total of 777 statements were recorded which translates into an average of 38.85 thoughts per subject. The two technologies, commercial aviation and atomic power, generated a total of 267 thoughts or an average of 13.35 thoughts per subject. Hazard elated thoughts accounted for about 80% of the total thoughts in the case of the consumer products and for 67% of the total thoughts involving the technologies suggesting that indeed the subjects tended to the instructions of the study and activated their hazard knowledge stored in memory.



The prototypical analysis of verbal protocols would now proceed to classify the reported thoughts without much attention given to the possibility of mental images having occurred during the generation of the protocols (Rossiter 1980). The information provided in Table 4, however, shows the potential seriousness of the neglect of imagery. It suggests that about 25-30% of the thoughts may have resulted from describing imagery occurred during the verbalizing task. On the average, every subject reported two images for the products and at least one image for the technologies.



The occurrence of imagery was further substantiated by the extent of imagery and vividness of imagery scales administered after the thought protocols, but before the directed image protocols. Subjects rated the extent to which images had entered their thoughts on a 7 point scale anchored by the descriptors "virtually none of my thoughts were in picture or image form" and "virtually all of my thoughts were in picture or image form." The average score across the consumer products was 5.40, with a low of 3.81 for household ammonia and a high of 5.91 for power lawn mower. The average for the two technologies was 4.10. Apparently, subjects at least think that some product knowledge is experienced in an image form. The vividness of the images experienced was rated on a 5 point categorical scale ranging from "perfectly clear and as vivid as actual experience" to "no images occurred, you only know you are thinking about hazard." The average vividness score for the reported images was 1.984 for the product images and 2.691 for the technology images. The most vivid images were associated bicycles (x = 1.28), power lawn mowers (1.62) and knives (1.71).

In summary, these data indicate that research on information representation may indeed be incomplete if it focuses exclusively on linguistic information. Images and the information contained in them are being activated by the subjects when considering the product and technology cues.

Content of Knowledge Structures

Table 5 explores the content of the hazard related thought generated in the thought protocols. The information indicates the predominant semanticity of the hazard related knowledge structures. Sixty nine percent of the thoughts could be characterized as semantic, i.e., as generic knowledge void of a personal reference. A significant portion, however, namely 25% of the thoughts did have some degree of personal reference. Personal episodic and episodic thoughts did account for 5 percent of the thoughts.



One aspect of the personal episodic and episodic thoughts not evidenced in the table should be reported. These two types of thoughts were typically extensively elaborated upon, far more so than in the case of the semantic thoughts. Sometimes entire paragraphs in the transcribed protocols were devoted to a description of a single episode in which the subject was a victim in an accident situation or a witness to such a situation. Semantic thoughts on the other hand were typically briefly stated and often expressed in terms of a listing of behaviors which people should not engage in so as to avoid specific hazards. In some cases this listing was quite long. Thus, the high percent of semantic thoughts should not be construed to indicate that attention to semantic memory to the exclusion of episodic memory is justified. OD the contrary, the mere fact that subjects reported past episodes in considerable detail may be thought of evidence of their importance. Indeed several of the episodes described had occurred in the distant past, such as childhood experience which further substantiates their importance.

The nature of the content of the reported images is portrayed in Table 6. For the consumer products the images are shown to be more personal experience based rather than neutral or induced by media report. The exception to this pattern is formed by Christmas light. A review of the protocols reveals that the image of a house-fire as a result of faulty Christmas lights as shown on the nightly news was a frequent media induced image which helps to explain this exception to the pattern. In contrast, the pattern in the images activated by the two technologies reveals a predominance of neutral and/or media induced images. These results seem to corroborate the findings by Combs and Slovic and Slovic et al. (1981).



Table 5 also indicates the spontaneous activation of what can best be characterized as affect thoughts. As noted earlier the affect thoughts frequently occurred as an initial reaction to the hazard cues. These data might very tentatively be interpreted as some support for Bower's (1981) associative network theory of memory and emotion.

In summary, the information provided in Tables 5 and 6 suggests that focus on hazard representations in "semantic memory" is justified on the one hand by the sheer volume of thoughts characterized as semantic. On the other hand, exclusive study of semantic memory would ignore a smaller but important content of memory. The data provide some indication that episodic thoughts and images of past hazard events are stored in memory and are being activated by a subject when required.


This exploratory study of representations of product hazards in consumer memory has generated as much speculation as confirmation.

The protocols generated by the elicitation procedure to show the importance of visual information in consumer hazard knowledge structures. This finding provides empirical evidence for Rossiter's (1980) admonishment of consumer researchers about the existing bias in consumer memory and research procedures toward linguistic information and verbally oriented measures. Indeed, some information has been provided to suggest the importance of visual information in such information processing activities as perception and inference drawing.

The exploration of the content of both the verbal and imagery protocols argues against the exclusive study of semantic memory. Episodically oriented thoughts and images have been shown to be stored in memory and may be influencing hazard perceptions. Indeed one of the more striking results of the study was that "personal episodic" and "episodic" verbal statements were recalled in extensive detail. Though a far greater number of semantic statements were reported, episodes relevant to the self or a significant other seemed to be very well recalled and highly integrated. The content of the mental images seemed to be even more dominated by experiences in which the subject was either a participant or a personally involved witness. These results may be viewed as being directly opposite to Tulving's initial intent of showing weaker episodic memory than memory for abstract knowledge.

Such work remains to be done in the area of specifying how knowledge structures influence the processes of comprehension, perception, retrieval and storage. For example, for some of the products such as power lawnmowers and household ammonia as well as for atomic power a positive relationship was found between the number of images and perceived hazardousness. Interestingly, in the case of atomic power, the number of neutral/media induced images also showed significant correlation with the perceived hazardousness scales hinting at the role of the media in presenting the hazards associated with atomic power. Further empirical work on these relationship would have both theoretic and practical implications.

The examination of the content of the thoughts also raise auxiliary questions about structure or organization. The content and organization of knowledge structure are interrelated in as much as the strengths of the structural links between nodes are based upon the meaning of the concepts themselves. Further work on the exact nature of the interrelation between content and structure would help in refining the conceptual and operational definitions of schemata and their application in consumer research.


Abelson, Robert P. (1976), "Script Processing in Attitude Formation and Decision Making," (eds.) J. S. Carroll and J. W. Payne, Cognition and Social Behavior, NY: Lawrence Erlbaum.

Abelson, Robert P. (1980), "The Psychological Status of the Script Concept," Cognitive Science Technical Report #2, Yale University.

Anderson, John R., and Bower, Gordon R. (1973), Human Associative Memory, Washington: Winston.

Anderson, John R. and Ross, Brian H. (1980), "Evidence Against a Semantic-Episodic Distinction," Journal of Experimental Psychology: Human Learning and Memory, Vol. 6, No. 5, pp. 441-466.

Bobrow, Daniel G., and Norman, Donald A. (1975), "Some Principles of Memory Schemata," (eds.), D. G. Bobrow and A. M. Collins, Representation and Understanding: Studies in Cognitive Science, NY: Academic Press.

Brucks, Merrie, and Mitchell, Andrew A. (1981), "Knowledge Structures, Production Systems and Decision Strategies," (ed. ), Kent B. Monroe, Advances in Consumer Research, Vol. 8, Ann Arbor, MI: Association for Consumer Research, pp. 750-757.

Bower, Gordon H. (1981), "Mood and Memory," American Psychologist, (February) pp. 129-148.

Calder, Bobby J. (1978), "Cognitive Response, Imagery and Scripts: What is the Cognitive Basis of Attitude," (ed.), H. Keith Hunt, Advances in Consumer Research, Vol. 5, Ann Arbor, MI: Association for Consumer Research, pp. 630-634.

Collins, Allan M., and Loftus, Elizabeth F. (1975), "A Spreading Activation Theory of Semantic Processing," Psychological Review, 82, pp. 407-428.

Combs, Barbar, and Slovic, Paul (1979), "Biased Newspaper Coverage and Biased Judgments," Journalism Quarterly, 56, pp. 837-843.

Fischhoff, Barush, Slovic, Paul, and Lichtenstein, S-rah (1980), "Lay Foibles and Expert Fables in Judgments About Risk," to appear in T. O'Riordan and R. R. Turner (eds.), Progress in Resource Management and Environmental Planning, Vol. 3, Chichester: Wiley, (in press).

Fischhoff, Barush, Slovic, Paul, Lichtenstein, Sarah, Read, Stephen, and Combs, Barbara (1979), "How Safe is Safe Enough? A Psychometric Study of Attitudes Toward Technological Risks and Benefits," Policy Sciences, 9, pp. 128-152.

Hermann, Douglas J., and Harwood, James R. (1980), "More Evidence for the Existence of Separate Semantic and Episodic-Stores in Long-Term Memory," Journal-of Experimental Psychology: Human Learning and Memory, Vol. 6, No. 5, pp. 467-478.

Johnson, Eric J., and Russo, J. Edward (1978), "The Organization of Product Information in Memory Identified by Recall Times," (ed.), H. Keith Hunt, Advances in Consumer Research, Vol. 5, Ann Arbor, MI: Association for Consumer Research. pp. 79-86.

Kanwar, Rajesh, Olson, Jerry C., and Sims, Laura S. (1981), "Toward Conceptualizing and Measuring Cognitive Structures," (ed.), Kent B. Monroe, Advances in Consumer Research, Vol. 8, Ann Arbor, MI: Association for Consumer Research, pp. 122-127.

Kates, Robert W. (1962), Hazard and Choice PercePtion in Flood Plain Management, University of Chicago, Department of Geography, Research Monograph, No. 78.

Kintsch, Walter (1980), "Semantic Memory: A Tutorial," in R. S. Nickelson (ed.), Attention end Performance, Vol. 8 Hillsdale, NJ: Lawrence Erlbaum, pp. 595-619.

Lichtenstein, Sarah, Slovic, Paul, Fischhoff, Baruch, Layman, Mark and Combs, Barbara (1978), "Judged Frequency of Lethal Events," Journal of Experimental Psychology: Human Learning and Memory, Vol. 4, No. 6, pp. 551-578.

Olson, Jerry C. (1980), "Encoding Processes: Levels of Processing and Existing Knowledge Structures," in Jerry C. Olson, (ed.) Advances in Consumer Research, Vol. 7, Ann Arbor, MI: Association for Consumer Research, pp. 154-160.

Olson, Jerry C. (1981), "The Importance of Cognitive Processes and Existing Knowledge Structures for Understanding Food Acceptance," in Criteria for Foot Acceptance: How Man Chooses to Eat, (eds.) J. Soles and R. L. Hall, Zurich: Forster Publishing Company, pp. 69-81.

Olson, Jerry C, and Muderrisoglu, Aydin (1979), "The Stability of Responses Obtained by Free Elicitation: Implications for Measuring Attribute Salience and Memory Structure," in W. L. Wilkie, (ed.) Advances in Consumer Research, Vol. 6, Ann Arbor, MI: Association for Consumer Research, pp. 260-275.

Ortony, Andrew (1978), "Remembering, Understanding, and Representation," Cognitive Science, 2, pp. 53-69.

Otway, Harry J. (1973), "Risk Assessment and Social Choices" in Proceedings of the LAASA Planning Conference on Energy Systems, Laxamburg, Austria: International Institute for Applied Systems.

Paivio, Alan (1971), Imagery and Verbal Processes, New York: Rinehart and Winston.

Posner, Michael I. (1973), Cognition: An Introduction, Glenview, IL: Scott Foresman.

Rethans, Arno J. (1980), "Product Safety and Product Hazard Perceptions: Implications for Research and Public Policy," in R. P. Bagozzi et al., (eds.) 1980 Educators' Conference Proceedings, Chicago: American Marketing Association, pp. 428-431.

Rethans, Arno J. and Albaum, Gerald S. (1981), "Towards Determinants of Acceptable Risks," in Kent B. Monroe, (ed.) Advances in Consumer Research, Vol, 8, Ann Arbor, MI: Association for Consumer Research, pp. 506-510.

Richardson, Alan (1977), "Verbalizer-Visual: A Cognitive Style Dimension," Journal of Mental Imagery, 1 (Spring). pp. 96-110.

Rossiter, John R. (1976), "Visual and Verbal Memory in Children's Product Information Utilization," in B. B. Anderson, (ed.) A Rethans in Consumer Research, Vol. Chicago: Association for Consumer, pp. 523-527.

Rossiter, John R. (1980), "Representation of Information in Memory: Some General Issues," in Jerry C. Olson, (ed.) Advances in Consumer Research, Vol. 7, Ann Arbor, MI: Association for Consumer Research, pp . 429-430.

Rossiter, John R. and Percy, Larry (1979), "Visual Communications in Advertising," to appear in R. J. Harris, (ed.) Information Processing Research in Advertising, Hillsdale, NJ: lawrence Erlbaum, in press.

Russo, J. Edward, and Johnson, Eric J. (1980), "What Do Consumers Know About Familiar Products," in Jerry C. Olson, (ed.) Advances in Consumer Research, Vol. 7, Ann Arbor, MI: Association for Consumer Research. pp. 417-423.

Shepard, Roger N. (1978), "The Mental Image" American Psychologist, (February), pp. 125-137.

Slovic, Paul, Fischhoff, Barush and Lichtenstein, Sarah (1977), "Risk Assessment: Basic Issues," in R. W. Kates, (ed.) Managing Technological Hazard: Research Needs and Opportunities, Monograph #25, University of Colorado, Institute of Behavioral Science.

Slovic, Paul, Fischhoff, Barush and Lichtenstein, Sarah (1979), "Rating the Risks," Environment, 21 (April), pp. 14-39.

Slovic, Paul, Fischhoff, Barush and Lichtenstein, Sarah (1981), "Perceived Risk: Psychological Factors end Social Implications," Proceedings of the Royal Society London, A376, pp. 17-34.

Slovic, Paul, Fischhoff, Barush and Lichtenstein, Sarah (1981), "Facts and Fears: Societal Perception of Risk," in Kent B. Monroe, (ed.) Advances in Consumer Research, Vol. 8, Ann Arbor, MI: Association for Consumer Research. pp. 497-502.

Thomas, Kerry, Swaton, Elizabeth, Fishbein, Martin, and Otway, Harry, J. (1980), "Nuclear Energy: The Accuracy of Policy Makers' Perceptions of Public Beliefs," Behavior Science, Vol. 25. pp. 332-344.

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

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

Whites R., Sheehan, P. W., and Ashton, R. (1977), "Imagery Assessment: A Survey of Self Report Measures," Journal of Mental Imagery, 1, pp. 145-170.

Wickelgren, Wayne A. (1981), "Human Learning and Memory," Annual Review of Psychology, 32, pp. 21-52.

Zajonc, R. B. (1980), "Feeling and Thinking: Preferences Need No Inferences," American Psychologist, (February), pp. 151-175.