The Informational Basis of Self-Reports; a Preliminary Report

Peter Rip, University of Chicago
ABSTRACT - Self-reports about decision processes are a common source of data in consumer research. Nisbett and Wilson (1977) have recently renewed our interest in the accuracy of these reports. Several conceptual issues relevant to the study of self-reports on decision processes are reviewed. An experimental paradigm for studying the contents of self-reports and their relation to the processes by which reports are generated are discussed.
[ to cite ]:
Peter Rip (1980) ,"The Informational Basis of Self-Reports; a Preliminary Report", in NA - Advances in Consumer Research Volume 07, eds. Jerry C. Olson, Ann Abor, MI : Association for Consumer Research, Pages: 140-145.

Advances in Consumer Research Volume 7, 1980     Pages 140-145

THE INFORMATIONAL BASIS OF SELF-REPORTS; A PRELIMINARY REPORT

Peter Rip, University of Chicago

ABSTRACT -

Self-reports about decision processes are a common source of data in consumer research. Nisbett and Wilson (1977) have recently renewed our interest in the accuracy of these reports. Several conceptual issues relevant to the study of self-reports on decision processes are reviewed. An experimental paradigm for studying the contents of self-reports and their relation to the processes by which reports are generated are discussed.

1.0  WHY STUDY SELF-REPORTS?

Subjects in an experiment on decision making are asked to "think aloud" anything which comes to mind.

Immediately following a decision, subjects in some other experiment are asked to try and recall how they chose between alternatives.

As a participant in a survey of recent new car purchasers, you are asked to report how you selected your car.

A salesperson is trying to understand what you look for when purchasing and asks you to tell him how you select this type of product.

The responses to all these questions are self-reports on one's decision- making processes, i.e., statements which answer why certain responses were or are being made. Yet the questions are distinct. Their answers would vary considerably, though there may be a thread of resemblance running through them.

Self-reports are the most relevant type of datum collected in consumer research and have been used to measure a wide array of attitudes, intentions, internal processes or states with Likert scales, verbal protocols, focus-groups, diaries, and a host of various pencil and paper measures. In addition they have been used as reliable reports of past behaviors such as purchases. Although the technique may differ from measure to measure, all of these instruments are merely channels by which the subject communicates data about his experiences to the researcher. Given the widespread use of self-reports, it is important to consider the rationale for their use. We will address a special type of self-report: self-reports on decision-making processes. This is especially important since Nisbett and Wilson (1977) have recently rekindled an old controversy in the behavioral sciences concerning a person's ability to report on the causes of his/her own behaviors. The purpose of this paper is to examine the issue of accuracy of self-reports and, in so doing, to develop an experimental paradigm within a broad theoretical context. Some preliminary results of experiments conducted in this paradigm will be reported as well.

1.0  WHAT ARE SELF-REPORTS REPORTING ABOUT?

1.1  Goals Of Self-Reports

Two dimensions we might use to contrast different types of self-reports on choice processes might be recency since the decision, the time dimension, or reporting context (researcher's lab v. natural social settings). These distinctions capture some aspects of the differences among types of self-reports. Reliability and completeness might be other ways to contrast self-reports on decision-making processes. (Ericsson & Simon 1978). While these may be significant stinctions, they may not be the most significant bases for distinguishing between various types of self-reports. A more important criterion is the goal of the report. The goal will determine the process by which the report is generated, the cognitive and informational resources employed, the context of the report, and the types of environmental influences which shape the report, e.g. experimenter's probes, social motivations, etc.

1.1.1  Concurrent Reports. Concurrent protocols are a popular process-tracing technique in decision research. The goal of this procedure is an accurate reflection of the contents of short-term memory, thereby providing some insights into mental events. Any theoretical interest in these reports is motivated by their use as measurement instruments, not as naturally occurring behaviors, precisely because of the artificiality of thinking aloud.

1.1.2  Retrospections.  Retrospections are usually motivated by one of two goals. In recall measures the goal of the researcher is an unfiltered telling of the events which occurred. The subject's goal in either case is to provide reports which are complete and unbiased, assuming away social motivations. The research issue is the content of memory, rather than how this report is generated. The resources employed in the formation of retrospective reports are somewhat different from those in concurrent reports. The primary informational resource is more restricted, since fewer thoughts will be retained in a long-term memory. Processing resources are greater, because there is no concurrent task to perform. A second type of retrospective self-report on decision processes may be termed explanations. Explanations are not motivated by an unfiltered re-telling of events. The goal of explanation is to provide structure and coherence. Recollections are a foundation for explanations, but something is added. Explanations are syntheses of information. Filtering and selectivity are part of the generating process. Understanding is achieved through the imposition of structure, not necessarily through accuracy. Accuracy is an occasional by-product.

1.2  The Contents of Reports

We have not yet answered the opening question to this section: What are all these self-reports reporting about? The obvious response is that it depends on the question or probe put to the consumer (Ericsson & Simon 1978). More generally, we can distinguish at least three different types of information which can be reported: procedure, contents without reference to procedure, or reasons for attending to or using specific procedures or contents. The prevalence of each of these three subjects will differ as a function of the goal of the report.

1.2.1  Procedure .  How would I go about choosing a college? How am I selecting between these six products? How did I choose to go to this Italian restaurant tonight? The answer to these questions requires some abstraction or generalization over the set of choice alternatives. The response is a general rule or set of transformations which account for the ordering of preferences among the alternatives. The answer does not require a complete replaying of every operation performed on the path to a preference. Only the general transformational rule or strategy is required. This is the question of procedure and one which decision researchers have come to refer to as choice rules or strategies [Note that answers to these sorts of questions may vary in levels of response. "I chose the one which is most like my friends'" and "I chose the one which survived the following six cutoffs" may be isomorphic rules, although at different levels of response. None of the responses are incorrect, though some may be more complete than others.].

1.2.2  Contents.  It is possible for consumers to report contents without reference to any process. The simplest examples would be statements such as "I like the price" or "I noticed this one was very large." These statements are contrasted with remarks such as "I first eliminated all those costing more than $100." Reports of content are statements of fact, perception, belief. They are symbols which are manipulated by cognitive procedures (Newell & Simon 1972).

1.2.3  Rationale.  A third topic which may be present in a report is one's rationale. A decision maker may have a rationale for either processes or contents, i.e. "I first checked the prices because..." or "I like large cars because..." Reasons generally contain references to goals or sub-goals. They are the motivational elements in decision making. Reasons are not causal. They explain procedures or contents, but not preferences. Therefore, they are ancillary issues in studying the quality of self-reports as causal models of one's choices.

As mentioned earlier, the prevalence of each of these three subjects, rationale, procedures, and contents, or why, how, and what, may differ across types of self-reports. Concurrent protocols may reflect a great deal of "what" content, i.e., statements about which characteristics the consumer is attending to, couched with some "why" and "how". For example, a consumer thinking aloud may say something such as "I see this one is economy size. I guess I will buy it since it has the lowest unit price and that saves me money in the long run." As a general observation, retrospective reports may emphasize the how and why over the what.

2.0  HOW CAN SELF-REPORTS BE INACCURATE?

2.1  Methods Of Studying Accuracy

Given the potential subjects of self-reports, how can self-reports be inaccurate? An interest in their accuracy presupposes some unequivocal standard against which to compare the reports. This can be problematic, since we as researchers take the reports as surrogates for states we could not otherwise observe. This leaves us with two alternatives in developing standards against which to compare the reports,

One approach is to create a task which has a simple structure, though the structure may be hidden to the consumer. The general model is to create a situation in which there is only one path from A to B. All those who arrive at B are then questioned about the path, Accuracy is defined as citation of the path as constructed by the researcher. The procedure assumes that the psychological representation of the task, the problem space (Newell & Simon 1972), follows the physical representation.

While this first case corresponds to the experimental model of research where all elements but one are held constant, there is a modeling alternative which is the regression approach. It, too, requires a strong assumption; an assumption about what the path from A to B looks like. The path is usually a paramorphic model of the judge and the parameters of that model are estimated statistically. Decision makers are asked questions which correspond to parameters of that model. Accuracy of the report is defined as accurate parameter estimation. The null hypothesis of accurate reports of parameter values is a strong hypothesis, i.e., an exact prediction. Therefore, it is easy to statistically reject this hypothesis of accuracy of report. There is also a potential problem of specification error; the model form may be inaccurate, not the consumer's report. This only affects the researcher's ability to estimate the degree of accuracy. It does not affect our ability to examine how accuracy is affected by other variables such as task structure or individual differences, unless those variables simultaneously affect the "true" model used by consumers.

2.2  Types of Errors

2.2.1  Errors Of Transmission.  One potential source of inaccuracy in self-reports has nothing to do with whether the consumer understands his/her decision making. This is an error in communication. Errors in transmission may fall into two general classes. One such error arises from the respondent's lack of adequate vocabulary to describe choice processes [This suggestion is due to Lee G. Cooper.]. This may be due to lack of experience or training. The second error is a levels -of- response problem. The response given may be merely a translation of the true response, but is not recognized as such. Neither of these "errors" are errors in the reports themselves. They are errors in the procedure. As researchers we have responsibility for 1) establishing that the decision makers we study have the ability to respond effectively, and 2) ensuring that respondents understand the criteria which will be used to judge their accuracy.

2.2.2  Errors Of Commission.  Assume that the above source of communication problems are ruled out. How else may self-reports be inaccurate? There may be systematic deviations between the "true" standard and the self-report. The consumer may misspecify the procedure, the goals, or the relevant contents to which he attends. He may report having been lexicographic on a dimension, when in fact he was not. He may report eliminating from consideration a group of products because his wife would not like them, when the real reason was because they were not the style in which he wanted to be seen. He may report examining the price of a product and finding it to be $4.95, when in fact the price was not given. Each of these are clear violations of the true how, why, or what.

Assuming that consumers are not out to deceive us, how do errors of commission arise? This must involve a lack of awareness of the true response. However, lack of awareness is not enough. There must be other mechanisms which account for an erroneous response in the face of no awareness. It depends on whether the decision maker is aware of his unawareness. I may be fully cognizant of my lack of awareness. In this case, the error of commission is explained by some social factors which force me to provide a response, fully aware I do not have the answer. The more interesting case is when I am unaware of my unawareness. The question remains as to where answers come from in these cases. We shall postpone this issue of the source of systematic misreporting for a brief while. These second errors of commission might be expected to be more resistant in the face of disconfirming evidence, because they are more strongly held to begin with. Awareness is central in either case.

2.2.2  Errors Of Omission.  In addition to being systematically biased in the information reported, it is also possible to err in self-reports by being incomplete. Incompleteness and bias are certainly related, since each involve a basic inability to report the entire "true" response. The role of awareness is the distinction between biased and incomplete responses. In the case of misreporting (errors of commission), the consumer does not know that he does not know.

When we speak of incompleteness, we must specify with regard to what. It is possible to be comprehensive, but distort particular elements. For example, suppose price is the first factor you screen on in making a decision and then you screen on size. If you report that you used two factors, price and size to arrive at your decision, this understates (misrepresents) their relative importance. Yet the report may be considered complete with respect to contents, since price and size are certainly included in a prominent position in the report.

The interesting issues in the study of completeness of self-reports are the conditions under which they will he more or less complete and whether types of information are systematically omitted. If this is to be addressed theoretically, we must begin with a model of awareness and attention to cognitive processes. A complete model of attention to cognitive processes would at least need to specify how motivation, attentional capacity, memory, and task environment interact to affect awareness and completeness. Ericsson and Simon (1978) have recently made some progress in this regard by applying Newell and Simon's (1972) information processing model to this problem. They have limited their analyses to self-reports as measures whose goals are unfiltered transmission of the contents of short- and long-term memory.

We need a theory of information in addition to a theory of awareness and attention. Procedure, reason, and contents, especially procedure and contents, are appealing candidates, but there are some important problems, as we shall soon see. A theory of information or knowledge to which decision makers may attend while choosing between alternatives is an important first step toward understanding systematic biases in attention and eventual self-reports. The remainder of this paper will consider what awareness is and lead toward an operationalization of a theory of information relevant to self-reports on consumers' decision making.

3.0  AWARENESS AND SELF-REPORTS

3.1  The Radical Position

Nisbett and Wilson (1977) recently restated a theoretical position which has its roots in Behaviorism (Skinner, 1974) and the philosophy of mind (Ryle, 1948). The argument resembles Bem's (1972) statement of self-perception theory, though this time applying the principles to self-reports of decision making processes rather than attitudes. Briefly, Nisbett and Wilson (1977) contend that no such introspective awareness exists. They hypothesize that self-reports are nothing more than a priori causal theories, not reports based on true self-awareness. We act as observers of our decisions and offer as self-reports the reports that anyone would give to account for our decisions.

This position against privileged access is not without critics. Smith and Miller (1978) provide a review of the issues surrounding this controversy. Their review of Nisbett and Wilson (1977) suggests that the decision makers have been misled. They argue that Nisbett and Wilson are "implicitly using an impossible criterion of introspective awareness: that subjects [are asked to report and] be aware of what we systematically and effectively hide from them by our experimental settings" (Smith & Miller, 1978, p. 356).

Smith and Miller (1978) also argue that the hypothesis of a priori theories is inherently non-falsifiable. This non-falsifiability stems from the fact that accuracy of self-insight is seen by Nisbett and Wilson as spurious correspondence between a priori theory and true cause or process. The spurious relationship stems from the random convergence of the true environmental cause and subjects' a priori beliefs and not correspondence of causes and introspections. Both of these problems arise from Nisbett and Wilson's (1977) pursuit of their Behaviorist position and apply equally well to the self-perception position advocated by Bem (1972).

3.2  Awareness As Available Knowledge

Awareness is a major problem in the Nisbett and Wilson (1977) model. It is a term left undefined, though critical to any study of self-reports on decision making. Their implicit definition of awareness of causal stimuli in self-reports is the natural tendency to cite the cognitive processes which resulted in a decision or action. They fail to distinguish between tendency and ability. It is possible that the appropriate causal knowledge is potentially available in memory, but currently inaccessible. Tulving and Pearlstone (1966) have demonstrated the importance of the distinction between potential availability of knowledge and its accessibility. They showed that when subjects were given proper cues, semantic recall could be enhanced considerably. Words which were otherwise inaccessible, due to their organization in memory, were made accessible through the use of cues which were similar to the category names on the original lists memorized by subjects. Similarly, Russo and Wisher (1976) have demonstrated that in simple arithmetic tasks, intermediate results of cognitive operations could be made more accessible retrospectively through the use of proper cues. The cues used by Russo and Wisher (1976) were traces of intermediate results which subjects obtained in the task. The conclusion must be that merely showing that decision makers do not naturally utilize introspective knowledge is not sufficient evidence to conclude no potential introspective awareness exists. Under specific task conditions is may exist. I may not be able to recall what I did, but I'll know it when I see it.

3.3  Process And Content

This awareness controversy rests upon a distinction between two general types of knowledge: process and content knowledge. Nisbett and Wilson's model is that contents are given in the task and available to everyone. Processes are inferred from contents and prior theories. Content knowledge corresponds to what we have labeled the "what", i.e. the object(s) of one's immediate attention. Process knowledge is less well-defined, though it must correspond to how decisions are made. Munsat (1972), using osmosis as an example, defines a process as a transformation over successive states from initial to final state. It is a transformation which yields intermediate results. This illustrates the difficulty of defining process and content knowledge. Processes cannot be separated from content. Processes, whether they are decision strategies or more general cognitive processes, are only constructs. They are inferred by both actor and observer by the observation of a series of changes of contents. Therefore, processes rarely appear without reference to specific content; however, contents can and often do appear without reference to processes.

It is this inequality of status between process and content knowledge in decision making which leads Smith and Miller (1978) to claim that there is no operational or logical difference between a decision maker's citing a process or merely its intermediate results. Presumably they would claim that the difference is only due to an error of transmission; decision makers have no label to apply to this series of transformations. However, this only blurs the distinction between process and content. In so doing, it limits our ability to generalize about the types of information to which consumers and other decision makers will attend and recall. There is a difference between the consumer who can report all the intermediate results and the one who can abstract himself from local attention and report general consistencies he followed.

3.4  Outstanding Issues

Possibly the most important contribution by Nisbett and Wilson (1977) is their suggestion that certain self-reports on decision making processes are judgments. When is a self-report more like a judgment or problem solving task than simply a replay of one's thoughts? A judgment representation of the report is probably more appropriate when the goal of the report is something other than unfiltered communication of contents of memory. This generally will place us in the realm of retrospective reports, though not always.

Others have suggested social goals or motivations may operate in certain self-report situations (Smith & Miller 1978; Weitz & Wright 1979). The desire to appear rational to others or to oneself may be one factor which affects how prior theories are tempered with one's recollections. This motivation follows from a general primary goal which is to provide coherence and structure in one's explanation, rather than accuracy. The explanation of one's own behavior to either researchers or friends involves a synthesis of two types of information: prior knowledge or "theories" and introspection.

Nisbett and Wilson (1977) claim there is no sample information (introspective awareness); hence, the prior theory equals the posterior report. The juxtaposition of this Bayesian representation of these self-reports-as-judgments and the psychology of self-reporting does raise several interesting questions. Three major research areas here are:

1. The mechanisms which generate prior theories,

2. The process of introspecting (cf. Ericsson & Simon 1978), and

3. The means by which prior theory and introspection are integrated to yield a self-report.

Especially interesting is the psychology of how conflicts between prior theories and introspections will be resolved and under which conditions. This formal model of reporting on decision making as an information integration task holds promise for the study of self-reports. However, we should recognize that it does not apply to all self-reports. For example, in concurrent protocols the goal is not synthesis and coherence, but unfiltered communication. An information processing model such as that developed by Ericsson and Simon (1978) may be more appropriate to investigate issues such as task and experimenter effects on completeness of reporting.

A Bayesian model is not necessary to generate the hypothesized result that priors equal posteriors. Prior theories can be viewed as hypotheses. Decision makers appear to seek confirming rather than disconfirming evidence for their hypotheses (see Einhorn & Hogarth, 1978, for an important discussion of this point). Therefore, it follows that prior theories would equal the report, since the hypotheses have not been disconfirmed. The difference between this hypothesis testing representation and the Bayesian one is that we admit that introspection exists, unlike Nisbett and Wilson (1977), but that its search is biased. This result is consistent with the memory results of Tulving and Pearlstone (1966) and Russo and Wisher (1976) cited earlier.

3.4.1  Process/Content Refined.  Much of the ambiguity in using this dichotomy as a general distinction is due to the need to define potential information in self-reports in an abstract manner. The extreme alternative is to define information at its most structural level, the information contained in the task, e.g. prices, size, colors, etc. This is also an inadequate starting point in the definition of information, because it lacks generality. One middle ground is to take one specific task and define the types of information potentially available to consumers. hen we can consider how representative this task is of the range of consumer decisions.

4.0  AN EXPERIMENTAL PARADIGM FOR STUDYING SELF-REPORTS

The task chosen for consideration is the familiar multi-alternative, multi-cue profile evaluation procedure (Wright & Rip, 1978). This task is appealing to consumer researchers because of its similarity to a variety of choice problems. It is appealing to us both for this reason and because it is well-structured. The importance of this structure will become clear as we consider the theoretical issues involved in explicating some potential types of knowledge to which consumers may attend in the preference decision.

4.1  The Problem Space

What we seek to do is define the psychological environment constructed by consumers in this task. The problem space is essentially the mental representation of the decision formulated by the consumer. The four major determinants of the structure of this space are:

1.  The information processing abilities of the consumer,

2.  The inherent structure of the task,

3.  The constraints and resources present in the environment,

4.  The prior knowledge which the consumer brings to the task (Newell & Simon 1972).

The hypothesis is that the more attention is paid to a particular element in the problem space, the more prominent it will be in his reports. We would also hypothesize that the accuracy of retrospective reports about the problem space, such as reports of its content or processes applied in it, would be related to attention to and retention of that space.

Our goal in empirical work will be to compare across consumers' problem space representations to develop a general statement about these representations. Several important assumptions will be required to develop a general problem space representation of this Judgment task. First, the types of knowledge consumers bring to these tasks is essentially equal across subjects. This assumption is probably met in practice by the use of hypothetical profiles, though one could make a case that such things as perceived covariation between attributes varies considerably across individuals. Second, we must assume all subjects have are given identical initial representations of the decision. This is achieved through the use of standardized instructions which provide some structure to the task. The instructions given to subjects are likely to have a major impact on the subjects' problem space representations (Ericsson & Simon 1978; Simon 1975; 1976).

The third and final assumption is that decision makers are equal in their information processing capabilities. While this is certainly implausible across the population, the assumption simplifies our theoretical task considerably by dismissing the significance of individual differences. Operationally this assumption requires considerable homogeneity in the decision makers under study. It becomes more plausible if we limit our subject pool to those who are more or less equal in motivation and familiarity with the problem, are similar in traits we might assume to be correlated with information processing capability such as verbal/quantitative intelligence, and are roughly at the same stage of cognitive development and experience. The plausibility of our discussion of the problem space representations constructed by decision makers faced with this evaluation task the validity of the assumptions. We will use this problem space to define which types of information ~ be available to decision makers.

The discussion has been motivated by an interest in the information upon which self-reports, especially retrospective self-reports, are based. One source of information is whatever thoughts are retained after the task. These thoughts are not random; they are generated in the process of developing and solving a problem space representation. To understand the types of information to which decision makers attend, we must understand the problem space which they construct. We now turn to a consideration of a problem space for a single task - evaluation of hypothetical profiles.

The bare structure of this task suggests four different components of the problem space. The first two components define the dimensionality of the problem space. These are 1) the number of alternatives or profiles, and 2) the dimensionality of the alternatives themselves, the dimensions or characteristics of the alternatives. Theoretically, the decision makers may attend to something about the dimensionality of the problem space by thinking about either of these two determinants. Examples of thoughts about the number of alternatives would be those where the decision maker thinks about attending only to a subset of the products or may consider how many of them are appealing for potential evaluation, or recalls he can only afford a certain to purchase a certain number. Other attempts to address the dimensionality of the problem space would be to effectively reduce the number of dimensions to which he/she will attend by thinking about how relatively or absolutely important the dimensions are, how correlated they are, etc. They are examples of how the decision maker can affect the dimensionality of the problem space via explicit generation of these two types of thoughts.

The individual attribute levels of the alternatives comprise a third subject of the decision maker's attention. We can either think of these as either the part-worth utilities of the attributes of the products or simply the relationships between these attributes and outcomes. No distinction is made at this point between these two types of thoughts. This is because either type of thought takes as its focus something about the attribute level itself. This thought may be either physical outcome or value. Of course at a higher level, both events and values associated with attributes are outcomes; one being material, the other being psychological. The essential point is that the thought is a direct reference to something about the attribute. This is most directly what Nisbett and Wilson (1977) and others refer to as "content" knowledge. These are thoughts about the specific data of the alternatives.

This final part of the problem solving involved in choice preference formation is what has come to be referred to as the strategy or rules. Strategy is something to which decision makers may attend as they form preferences, and something which they may reference in their self-reports. The strategy need not be the global type of lexicography, conjunction, etc. They may be more local considerations of logic being used, rules which of which the decision maker is aware or is trying to impose, or the process tracing which the decision maker uses to induct his/her own rules. The awareness of these abstract logical, conditional operations is the essence of these thoughts. It is not merely their application. Behaving in a manner consistent with a rule is not sufficient evidence for the awareness of either that consistency or the awareness of the existence of any rule. We will refer to these types of thoughts as strategy thoughts.

We can now turn to the issue of considering how much attention decision makers devote to the components of "process/structural knowledge" versus "content knowledge". There is one simple hypothesis with which we can begin; decision makers think much more about attributes than the other three topics. We can develop many reasons for this hypothesis, the simplest being that attributes are concrete and dominate our field of vision. The other three subjects are more abstract and require some degree of reallocation to elements which are not natural objects of attention. We might also add that these contents are the stuff on which our strategies (conscious or otherwise) will operate; therefore, they become the subjects of thought as we search and compare attributes.

5.0  TOWARD EMPIRICAL RESEARCH

Recent work using this typology in a concept evaluation task has shown some promise. The task was to evaluate hypothetical colleges (16 profiles, 5 cues each) with respect to probability of applying to them. The consumers were second semester high school juniors intending to apply to college the next year. A full 73.1% of decision makers' concurrent vocalizations were classified into at least one of the four categories outlined here. Eighty-three percent of their written retrospections were similarly classified. These studies also strongly support the hypothesis that attention and retention are dominated by thoughts about content rather than thoughts about strategy or structure. Using several indices computed from the reports, content dominated decision makers' attention and retention. Furthermore, the composition of their unfiltered reports was related to the degree of accuracy they showed in later judgments about their decision making processes.

Several areas appear promising for future research on how consumers develop causal reports about their own or others' behaviors. We have already considered attention in the task, retention, and the development of the problem space as further topics for research, especially with regard to relevant situational, task, and motivational factors. Equally as appealing is the study of how decision makers call upon prior theories and experience. What conditions must prevail for the decision maker to call upon one theory over another? Ultimately it may rest upon notions of "similarity" and and "salience". Alternatively, if one believes that prior theories are hypotheses in search of confirmation, then how are these hypotheses formed and how do they bias our introspections? Finally, how are these prior theories and one's recollections/introspections integrated? What does one do when the prior theory conflicts with recollections? What determines the diagnosticity of one's own remembrances? What are self-reports in social situations? How do the "reports" we develop for ourselves affect our ability to learn from outcomes of our decisions? The development of self-reports on decision making processes promises to be an interesting area for future research.

REFERENCES

Bem, D.J, Self-Perception Theory, In L. Berkowitz (Ed.), Advances in Experimental Social Psychology 6, New York: Academic Press, 1972.

Einhorn, H. J., Hogarth, P. M., Confidence in Judgment: Persistence of the Illusion of Validity, Psychological Review, 1978, 85, 395-416.

Ericsson, K.A., & Simon, H.A., Retrospective Verbal Reports as Data, unpublished manuscript, Carnegie-Mellon University C.I.P Working Paper No. 388, August. 1978.

Munsat, S., What is a Process?, American Philosophical Quarterly, 6, 1969, 79-83.

Newell, A., & Simon, H.A., Human Problem Solving, Englewood Cliffs, NJ: Prentice Hall, 1972.

Nisbett, R. E., & Wilson, T.D., Telling More Than We Can Know: Verbal Report on Mental Processes, Psychological Review, 1977, 84, 231-259.

Rip. P., The Extent and Basis of Insight in Decision Making: A Study in Consumer Behavior, Stanford University, unpublished dissertation, June, 1979.

Russo, J. E., and Wisher, R. A., Reprocessing as a Recognition Cue, Memory and Cognition, 1976, 4, 683-689.

Ryle, G., The Concept of Mind, New York: Barnes & Noble, Inc., 1949.

Simon, H. A., Information Processing Theory of Human Problem Solving, C.I.P. Working Paper No. 324, Carnegie-Mellon University, May 1976.

Simon, H. A., The Functional Equivalence of Problem Solving Skills, Cognitive Psychology, 1975, 7, 268-288.

Simon, H. A., & Newell, A., Thinking Processes. In D. H. Krantz, R.C. Atkinson, R. D. Luce, & P. Suppes, Learning, Memory, and Thinking, San Francisco: W.H. Freeman and Company, 1974.

Skinner, B.F., About Behaviorism, New York: A. Knopf, Inc., 1974.

Smith, E. R., & Miller, F. D., Limits on the Perception of Cognitive Processes: A Reply to Nisbett and Wilson, Psychological Review, 1978, 85, 355-362.

Tulving, E., & Pearlstone, Z., Availability Versus Accessibility of Information in Memory for Words, Journal of Verbal Learning and Verbal Behavior, 1966, 5, 381-391.

Weitz, B. & Wright, P., Retrospective Self-Insight About the Factors Considered in Product Evaluation, Journal of Consumer Research, 1979, in press.

Wright, P., & Rip, P., The Use of the Active Evaluation Paradigm in Studying Judgment Processes, in H. K. Hunt (ed.) Advances in Consumer Research 5, Ann Arbor: Assoc. for Consumer Research, 1978.

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