Constructive Consumer Choice Processes (Bettman Luce Payne)

Constructive Consumer Choice Processes
Author(s): James R. Bettman, Mary Frances Luce, John W. Payne

Source: The Journal of Consumer Research, Vol. 25, No. 3 (Dec., 1998), pp. 187-217
Published by: The University of Chicago Press
Stable URL: http://www.jstor.org/stable/2489831
Accessed: 22/04/2010 19:36
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=ucpress.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact [email protected].
The University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access to The
Journal of Consumer Research.
http://www.jstor.org
Constructive Consumer Choice Processes
JAMES R. BETTMAN
MARYFRANCESLUCE
JOHN W. PAYNE*
Consumer decision making has been a focal interest in consumer research, and
consideration of current marketplace trends (e.g., technological change, an infor-
mation explosion) indicates that this topic will continue to be critically important.
We argue that consumer choice is inherently constructive. Due to limited pro-
cessing capacity, consumers often do not have well-defined existing preferences,
but construct them using a variety of strategies contingent on task demands.
After describing constructive choice, consumer decision tasks, and decision strat-
egies, we provide an integrative framework for understanding constructive choice,
review evidence for constructive consumer choice in light of that framework, and
identify knowledge gaps that suggest opportunities for additional research.
Consumer choices concerning the selection, consump- defined preferences that do not depend on particularde-
tion, and disposal of products and services can often scriptions of the options or on the specific methods used
be difficult and are importantto the consumer, to market- to elicit those preferences. Each option in a choice set is
ers, and to policy makers. As a result, the study of con- assumed to have a utility, or subjective value, that de-
sumer decision processes has been a focal interest in con- pends only on the option. Finally, it is assumed that the
sumer behavior for over 30 years (e.g., Bettman 1979; consumer has ability or skill in computation that enables
Hansen 1972; Howard and Sheth 1969; Nicosia 1966). the calculation of which option will maximize his or her
One can infer from recent trends in the nature and struc- received value and selects accordingly. This approach to
ture of the marketplacethat the importanceof understand- studying consumer decisions, often attributed to econo-
ing consumer decision making is likely to continue. Rapid mists and called rational choice theory, has contributed
technological change, for instance, has led to multitudes greatly to the prediction of consumer decisions.
of new products and decreased product lifetimes. In addi- Over the past twenty-five years, an alternative, infor-
tion, new communications media such as the World Wide mation-processing approach to the study of consumer
Web have made enormous amounts of information on choice (e.g., Bettman 1979) has argued that rational
options potentially available (Alba et al. 1997). Further, choice theory is incomplete and/or flawed as an approach
consumers are often asked to make difficult value trade- for understanding how consumers actually make deci-
offs, such as price versus safety in purchasing an automo- sions. The information-processing approach endorses
bile, environmental protection versus convenience in a bounded rationality (Simon 1955), the notion that deci-
variety of goods, and quality of life versus longevity in sion makers have limitations on their capacity for pro-
complex health care decisions. cessing information. Such limitations include limited
How do consumers cope with the decisions they must working memory and limited computational capabilities.
make, some of which involve difficult trade-offs and un- In addition, decision makers are characterizedby percep-
certainties?One approachto studying consumer decisions tions attuned to changes ratherthan absolute magnitudes
has been to assume a rational decision maker with well- and diminishing sensitivity to changes to stimuli (Tversky
and Kahneman 1991 ). More generally, behavior is shaped
by the interaction between the properties of the human
information-processing system and the properties of task
*James R. Bettman is the Burlington IndustriesProfessor of Business environments (Simon 1990) .'
Administrationat the Fuqua School of Business, Duke University, Dur-
ham, NC 27708-0120. Mary Frances Luce is an assistant professor at
the Wharton School of the University of Pennsylvania. John W. Payne
is the Joseph J. Ruvane, Jr., Professor of Business Administration at 'For a recent comparison of these alternativeperspectives on decision
the Fuqua School of Business, Duke University. This research was making, see McFadden (1997). As McFadden notes, economists vary
supportedby grants from the Decision, Risk, and Management Science in terms of the degree to which they believe rational choice theory
Programof the National Science Foundation and from the Environmen- describes actual decision behavior. A middle ground adopted by some
tal Protection Agency. economists is that individuals have a consistent set of preferences but
that such preferences become known by the individual (are "discov-
187
C 1998 by JOURNAL OF CONSUMER RESEARCH, Inc. s Vol. 25 0 December 1998
All rights reserved. 0093-5301/99/2503-0001$03.00
188 JOURNAL OF CONSUMER RESEARCH
The notions of bounded rationality and limited pro- sometimes order another dessert. The probability that
cessing capacity are consistent with the growing belief prior preferences will be retrieved from memory and used
among decision researchers that preferences for options will depend on their relative accessibility and diagnos-
of any complexity or novelty are often constructed, not ticity, among other factors (Feldman and Lynch 1988).
merely revealed, in making a decision (Bettman 1979; Also, preferences may be more constructive to the degree
Bettman and Park 1980; Payne, Bettman, and Johnson that the decision problem is complex or stressful.
1992; Slovic 1995; Tversky, Sattath, and Slovic 1988). As noted above, an important implication of the con-
People often do not have well-defined preferences; in- structive nature of preferences (and evidence for such
stead, they may construct them on the spot when needed, construction) is that choices often are highly contingent
such as when they must make a choice. Thus, consumer on a variety of factors characterizing decision problems,
preference formation may be more like architecture, individuals, and the social context. For example, the fol-
building some defensible set of values, rather than like lowing are some of the major conclusions from research
archaeology, uncovering values that are already there on consumer decision making: (1) Choice among options
(Gregory, Lichtenstein, and Slovic 1993). depends critically on the goals of the decision maker. The
The idea of constructive preferences denies that indi- option that is selected will depend on the extent to which
viduals simply refer to a master list of preferences in the consumer's goals are minimizing the cognitive effort
memory when making a choice and also asserts that pref- required for making a choice, maximizing the accuracy
erences are not necessarily generated by applying some of the decision, minimizing the experience of negative
invariant algorithm such as a weighted adding model emotion during decision making, maximizing the ease of
(Tversky et al. 1988). Ratherthan one invariantapproach justifying the decision, or some combination of such
to solving choice problems, consumers appear to utilize goals. (2) Choice among options depends on the com-
a wide variety of approaches,often developed on the spot. plexity of the decision task. Options that are superior on
Consumers may also develop problem representationson the most prominent attribute are favored as the task be-
the spot by structuringor restructuringthe available infor- comes more complex because the use of simple decision
mation (Coupey 1994). One important property of this processes increases with task complexity. (3) Choice
constructive viewpoint is that preferences will often be among options is context dependent. The relative value
highly context dependent. In addition, because decision of an option depends not only on characteristics of that
approaches are developed on the fly, processing will be option but also on the characteristics of other options in
highly sensitive to the local problem structure.This im- the choice set. (4) Choice among options depends on
plies that processing approaches may change as consum- how one is asked; strategically equivalent methods for
ers learn more about problem structureduring the course eliciting preferences can lead to systematically different
of making a decision. decisions. (5) Choice among options depends on how the
Why are preferences constructive? One reason individ- choice set is represented (framed) or displayed, even
uals may constructpreferences is that they lack the cogni- when the representationswould be regardedas equivalent
tive resources to generate well-defined preferences for by the decision maker on reflection. A key issue in fram-
many situations (March 1978). A second importantrea- ing is whether the outcomes are represented as gains or
son is that consumers often bring multiple goals to a given losses, with losses impacting decisions more than corre-
decision problem. sponding gains.
Preferences are not always constructed;people do have Thus, constructive processing generally implies contin-
firm and stable preferences for some objects. For example, gent choices. However, the fact that a choice is contingent
the first author has a well-defined and highly positive need not imply that the processing was constructive, that
value for chocolate. In such cases, consumers may simply is, developed on the spot. For example, a consumer may
retrieve these previously formed evaluations from mem- have a well-established, but contingent, preference for hot
ory and select the option with the highest evaluation (i.e., chocolate on a cold day and a cold soda on a warm day;
affect referral [Wright 1975]). such a preference is not constructive.
People are most likely to have well-articulated prefer- The constructive view of consumer decision making
ences when they are familiar and experienced with the raises fundamentaltheoretical issues. It is clearly not suf-
preference object, and rational choice theory may be most ficient to respond "it depends" when asked to describe
applicable in such situations. Even in such cases, how- consumer choices. A major purpose of this article, there-
ever, situationalfactors may intrude;although a consumer fore, is to provide a conceptual frameworkfor understand-
has a strong preference for chocolate, he or she may ing constructive consumer choice. This framework then
allows us to accomplish the two other major goals of the
article: (1) reviewing consumer decision research with
ered") only through thought and experience (Plott 1996). In a similar the framework serving as an organizing device, and (2)
vein, Lucas (1986, p. S402) has argued that economists tend "to focus
on situations in which the agent can be expected to 'know' or to have using this review to find gaps in our knowledge that sug-
learned the consequences of different actions so that his observed gest new research directions. The remainderof the article
choices reveal stable features of his underlying preferences." is structured as follows: We begin with a discussion of
CONSTRUCTIVE CONSUMER CHOICE PROCESSES 189
TABLE 1 about the values of the attributes,if there are more attri-
AN EXAMPLEOF A CONSUMER DECISION TASK butes that are difficult to trade off, and if the number of
shared attributesis smaller, among other factors.
Car Reliability Price Safety Horsepower We have presented some general properties of con-
sumer decision tasks above. One of the most important
A Worst Best Good Very poor findings from prior consumer research is that the same
B Best Worst Worst Good individual may use a variety of different strategies when
C Poor Very good Average Average
D Average Poor Best Worst
making decisions. A great deal of research has focused
E Worst Very poor Good Best on characterizingsuch strategies, their properties, and the
factors influencing their usage. In the next section we
NOTE.-Attributes are scored on seven-point scales rangingfrom best to provide a brief overview of consumer decision strategies
worst,withbest indicatingthe most desirablevalueforthe attributeand worst
indicatingthe least desirablevalue. and their properties. We examine the determinants of
strategy use in a later section.
the natureof consumer decision tasks and decision strate- Consumer Decision Strategies
gies as necessary backgroundfor presenting the proposed Characteristics of Decision Strategies. We begin by
framework. Then we present the conceptual framework, considering four primary aspects that characterizechoice
provide a selective review of the literature on consumer strategies: the total amount of information processed, the
decision making, and enumerate proposed areas for new selectivity in information processing, the pattern of pro-
research. cessing (whether by alternative [brand] or by attribute),
and whether the strategy is compensatory or noncompen-
CONSUMER DECISION TASKS AND satory.
DECISION STRATEGIES First, the amount of information processed can vary a
great deal. For example, an automobile choice may in-
Decision Tasks and the Consumer volve detailed consideration of much of the information
Information Environment available about each of the available cars, as implied by
most rational choice models, or it may entail only a cur-
A typical consumer choice, such as the simplified auto- sory consideration of a limited set of information (e.g.,
mobile choice task illustrated in Table 1, involves a set repeating what one chose last time).
of alternatives, each described by some attributesor con- Second, different amounts of information can be pro-
sequences. The set of alternatives can vary in size from cessed for each attribute or alternative (selective pro-
one choice to the next, with some choices involving as cessing), or the same amount of information can be pro-
few as two options and others potentially involving many cessed for each attribute or alternative (consistent
more (in some cases the two options may be simply to processing). For example, suppose a consumer consider-
either accept or reject an alternative). The attributesmay ing the cars in Table 1 decided that safety was the most
vary in their potential consequences, their desirability to important attribute, processed only that attribute, and
the consumer, and the consumer's willingness to trade chose car D, with the best value on that attribute. This
off less of one attributefor more of another.For example, choice process would involve highly selective processing
a consumer may be fairly certain about the values of some of attributeinformation (since the amount of information
of the attributes (e.g., horsepower) but more uncertain examined differs across attributes) but consistent pro-
about others (e.g., reliability). The consumer may not cessing of alternativebrand information (since one piece
have information for all of the options on some attributes of information is considered for each car). The fact that
(e.g., reliability information would not be available for a working memory capacity is limited effectively requires
new model). In addition, some attributes,such as safety, selective attention to information. In general, the more
may be difficult for consumers to trade off; making trade- selective consumers are in processing information, the
offs requires possibly accepting a loss on such an attri- more susceptible their decisions may be to factors that
bute, with potentially threatening consequences. Finally, influence the salience of information, some of which may
some choices may not correspond to the above example, be irrelevant.
in which all options are from the same product category. Third,informationmay be processed primarilyby alter-
Individuals can be faced with choices in which the attri- native, in which multiple attributes of a single option
butes defining the options may differ, such as deciding are processed before another option is considered, or by
whether to spend money on a vacation or on a new stereo. attribute,in which the values of several alternatives on a
Such choices have been called noncomparable (Bettman single attribute are examined before information on an-
and Sujan 1987; Johnson 1984). The difficulty of the other attribute is considered. For example, a consumer
choice problem faced by the consumer will increase with might engage in attribute processing by examining the
more options and attributes, with increased uncertainty price of each of the five cars, concluding that car B was
the most expensive, car A was the least expensive, and examine reliability (and no other information) for all five
that car C had a very good price. However, the consumer cars, and choose car B. The lexicographic strategy in-
could process in an alternative-basedfashion by examin- volves limited, attribute-based, noncompensatory pro-
ing the reliability, price, safety, and horsepower of car A cessing that is selective across attributes and consistent
in order to form an overall valuation of that car. Many across alternatives.
standardmodels of decision making (e.g., weighted add- Satisficing is a classic strategy in the decision-making
ing) assume alternative-basedprocessing, although attri- literature (Simon 1955). Alternatives are considered se-
bute-basedprocessing is often easier (Tversky 1972). quentially, in the order in which they occur in the choice
Finally, an importantdistinction among strategies is the set. The value of each attribute for the option currently
degree to which they are compensatory. A compensatory under consideration is considered to see whether it meets
strategyis one in which a good value on one attributecan a predeterminedcutoff level for that attribute.If any attri-
compensate for a poor value on another. A compensatory bute fails to meet the cutoff level, processing is terminated
strategythus requires explicit trade-offs among attributes. for that option, the option is rejected, and the next option
Deciding how much more one is willing to pay for very is considered. For example, car A might be eliminated
good ratherthan average reliability in a car involves mak- very rapidly because it has the worst level of reliability.
ing an explicit trade-off between reliability and price, for The first option that passes the cutoffs for all attributes
example. Frisch and Clemen (1994), among others, have is selected. If no option passes all the cutoffs, the levels
argued that making trade-offs is an important aspect of can be relaxed and the process repeated. One implication
high quality, rational decision making. In a noncompensa- of satisficing is that which option is chosen can be a
tory strategy, a good value on one attributecannot make function of the order in which the options are processed.
up for a poor value on another. If a consumer decides to The satisficing strategy is alternativebased, selective, and
choose the safest car, then car D will be chosen regardless noncompensatory. The extent of processing will vary de-
of its high price and regardless of the high ratings for car pending on the exact values of the cutoffs and attribute
B on reliability or car E for horsepower. levels.
Elimination-by-aspects (EBA) combines elements of
Specific Decision Strategies. There are many differ- both the lexicographic and satisficing strategies. Elimina-
ent decision strategies, and these strategies can be charac- tion-by-aspects eliminates options that do not meet a min-
terized by the above aspects of choice processing. One imum cutoff value for the most important attribute.This
classic decision strategy is the weighted adding strategy. elimination process is repeated for the second most im-
Assume that the consumer can assess the importance of portantattribute,with processing continuing until a single
each attributeand assign a subjective value to each possi- option remains (Tversky 1972). In our car example, sup-
ble attribute level. Then the weighted adding strategy pose that the consumer's two most important attributes
consists of considering one alternative at a time, examin- were reliability and safety, in that order, and that the
ing each of the attributesfor that option, multiplying each cutoff for each was an average value. This consumer
attribute's subjective value times its importance weight would first process reliability, eliminating any car with a
(e.g., multiplying the subjective value of average reliabil- below-average value (cars A, C, and E). Then the con-
ity in a car times the importance of a car's reliability), sumer would consider safety for cars B and D, eliminating
and summing these products across all of the attributes car B. Hence, car D would be selected. Elimination-by-
to obtain an overall value for each option. Then the alter- aspects is attributebased, noncompensatory, and the ex-
native with the highest value would be chosen. Weighted tensiveness and selectivity of processing will vary de-
adding is therefore characterizedby extensive, consistent pending on the exact pattern of elimination of options.
(not selective), alternative-based,and compensatory pro- The equal weight strategy, a variation on weighted add-
cessing. Because weighted adding is extensive, compen- ing, considers all of the alternativesand all of the attribute
satory, and involves explicit trade-offs, it is often consid- values for each alternative. However, processing is sim-
ered to be more normatively accurate than heuristics that plified by ignoring information about attribute weights.
do not possess these characteristics (Frisch and Clemen A value is obtained for each alternative by summing all
1994). Weighted adding, however, potentially places of the attributevalues for that option, and the alternative
great demands on consumers' working memory and com- with the highest value is selected. The equal weight strat-
putational capabilities. Nevertheless, weighted adding is egy is thus a special case of weighted adding if unit
the decision model that underlies many of the techniques weights are assumed. The equal weight strategy has often
used by market researchers to assess preferences. been advocated as a highly accurate simplification
The lexicographic strategy provides a good contrast to (Dawes 1979). Processing is extensive, consistent, alter-
weighted adding: the alternative with the best value on native based, and compensatory.
the most importantattributeis simply selected (assuming The majority of confirming dimensions strategy was
that there are no ties on this attribute). If a consumer first described by Russo and Dosher (1983). Alternatives
believed that reliability was the most important attribute are processed in pairs, with the values of the two alterna-
for cars, he or she could use a lexicographic strategy, tives compared on each attribute,and the alternative with
CONSTRUCTIVECONSUMER CHOICE PROCESSES 191
a majorityof winning (better) attributevalues is retained. TABLE 2

The retained option is then compared to the next alterna- CHARACTERISTICSOF DECISION STRATEGIES
tive from the choice set, and this process of pairwise
comparison continues until all the alternatives have been Amount of Selective (S) Attribute-based (AT)
evaluated and one option remains. This strategy is a sim- information versus versus alternative-
plified case of a more general model of choice, additive Strategy processed consistent (C) based (AL)
difference (see Tversky [1969] for a presentation of the
WADD Extensive C AT
additive difference model and a discussion of the relation- LEX Limited S AL
ship between additive difference and adding models). SAT Variable S AL
Processing is extensive, consistent, attribute based, and EBA Variable S AT
compensatory. EQW Extensive C AL
MCD Extensive C AT
Alba and Marmorstein(1987) proposed that consumers FRQ Variable Variable AL
may evaluate and choose alternatives using counts of the CCM Variable C Both AT, AL
numberof good or bad features characterizingthe alterna-
tives. Consumers develop cutoffs for specifying good and NOTE.-WADD= weightedadding;LEX= lexicographic;SAT= satisficing;
bad features, and depending on the consumer's focus (i.e., EBA = elimination-by-aspects;EQW= equal weight; MCD = majorityof
confirmingdimensions:FRQ= frequencyof good and/orbad features;CCM
good features, bad features, or both), different versions = componentialcontext model.
of the strategy could be developed. Note that this strategy
can be seen as a voting rule applied to multiattribute
choice, where each attribute has one vote. Weber,
Simonson and Tversky 1992). Finally, a consumer might
Goldstein, and Barlas (1995) provide evidence consistent
compare the relative advantages and disadvantages of an
with such a strategy, noting that encoding of such out-
option to those of other options (e.g., compromise effects;
comes is often simple. For example, whether an outcome
Simonson and Tversky 1992). For other work on rela-
is a gain or a loss may have a largerimpact on preferences
tional heuristics, see Drolet (1997).
than the actual magnitude of the outcome. The amount
Tversky and Simonson (1993) propose a model called
and selectivity of processing for this strategy, as well
as whether it is compensatory or noncompensatory, will the componential context model that includes these types
of relational heuristics as special cases. In specific, they
depend on the variant of the rule used. Processing is
alternative based for all variants, however. define
Consumers also use combinations of strategies. A typi-
cal combined strategy has an initial phase in which some VB(X, S) = , ivi(xi) + 0 , R(x, y),
i~=l yES
alternatives are eliminated and a second phase where the
remaining options are analyzed in more detail (Payne where VB(x, S) is the value of option x given a choice set
1976; see Beach's [1990] image theory for furtheranaly- S and backgroundcontext B, Pi is the weight of attributei,
sis of the role of elimination or screening processes in vi (xi) is the utility of the value xi of option x on attribute
decision making). One frequently observed strategy com- i, R (x, y) is the relative advantage of option x over option
bination is an initial use of EBA to reduce the choice set y, and 0 is the weight given to the relative advantage
to two or three options followed by a compensatory strat- component of the model. The relative advantage term is
egy such as weighted adding to select from among those obtained by taking utility differences between the two
remaining. options on each attributeand combining them. This model
The strategies presented thus far have been those most is compensatory, consistent, and has both alternative-
commonly addressed in consumer research; they range based and attribute-based components. It is difficult to
from weighted adding to more heuristic strategies such assess the amount of processing for this model; for the
as EBA that process information in a more selective and small problems typically used in this research stream
noncompensatoryfashion. However, recent work utilizing (e.g., three options and two dimensions), the relative
more perceptual approaches (e.g., Simonson and Tversky advantagecomponent may be assessed essentially percep-
1992) has suggested the use of decision heuristics that tually, with little effort. For larger problems where direct
are relational and perceptual in nature. Such heuristics perceptual assessment may not be feasible, the amount of
emphasize the ratings of a given option relative to other processing could be extensive, although simpler versions
alternatives.For example, one heuristic of this type might could certainly be specified (e.g., only examine the rela-
examine dominance relationships among pairs of alternative advantage term for the two attributeswith the largest
tives and use such information to make a choice (e.g., the weights). We will returnto issues in specifying the pro-
choice of an asymmetrically dominating option; Huber, cesses implied by this model. For another model of rela-
Payne, and Puto 1982). Another exemplar might assess tive advantage combining absolute and comparative com-
relative trade-offs between pairs of options on pairs of ponents, see Shafir, Osherson, and Smith (1989, 1993).
attributes, compare these trade-off rates, and use these Table 2 provides a summary of the properties of the
comparisons as inputs to a choice (e.g., trade-off contrast; strategies considered above. Note that one can make infer-
ences about the types of strategies consumers use by ob- by providing insights into different aspects of the decision
serving such characteristics as the amount, selectivity, process. In addition to integrating them, each approach
and degree of alternative-based versus attribute-based can be extended in important ways. For example, we
processing. propose that the goals considered in the accuracy-effort
We have argued above that consumer choices are con- approachbe augmentedby including goals for minimizing
structive and have briefly noted some of the major find- the experience of negative emotion during decision mak-
ings from research on consumer decision making. In the ing and for maximizing the ease with which a decision
next section we attempt to provide an integrated frame- can be justified. We propose extending the perceptual
work that can explain such contingent choice patterns. approach by considering some perceptual processes as
choice heuristics whose propertiesand use could be exam-
AN INTEGRATED FRAMEWORK FOR ined in a manner similar to other heuristics. We call our
integrated framework a choice goals framework. In the
CONSTRUCTIVE CHOICE PROCESSES next section we outline the general principles of that
Payne (1982) reviewed two major frameworks for un- framework.
derstanding contingent choice: a cost-benefit (accuracy-
effort) approach and a perceptual approach. The basic
premise of the accuracy-effort approachis that each deci- A Choice Goals Framework
sion strategy can be characterizedby its accuracy and the In this section we outline the basic postulates of our
effort it requires in any given situation. Decision makers frameworkfor understandingconstructivedecision making.
select strategies in a situation based on some compromise Similar to Bettman's (1979) outline for an information-
between the desire to make an accurate decision and the processing theory of consumer choice, we consider what
desire to minimize cognitive effort. Since the accuracy consumers are trying to accomplish (goals), what influ-
and effort characteristicsgenerally differ across strategies ences the informationthey attendto and how it is perceived
for a given decision environmentand across environments or encoded, and what factorsaffect the heuristicsconsumers
for a given strategy, strategy usage will vary depending utilize to combine information and make a decision. For
on the properties of the decision task. each topic, we summarizeour majorconclusions as proposi-
The perceptual framework is usually associated with tions; these propositions underlying our integratedframe-
the work of Tversky and Kahneman;they prefer explana- work are summarizedin Appendix A.
tions of constructive decision making based on principles
of human perception. For example, Kahneman and Tver- Consumer Goals. Choices are made to achieve goals.
sky (1979) argue that our perceptions are attunedto notic- Bettman (1979) instantiated this notion by postulating
ing changes rather than absolute magnitudes of stimuli that a consumer making a decision has a goal hierarchy,
and that outcomes will naturally be coded as gains and often developed constructively on the spot, specifying the
losses relative to some reference point. They also argue goals and subgoals he or she must attain. More recently,
that different ways of framing a problem may lead to interpretive research on consumer behavior has focused
different choices, much like the effect of taking different on what choices and consumer possessions mean to con-
perspectives on perception; that subjects are often un- sumers (Belk 1988). Thus, it is critical to characterize a
aware of the effects of framing; and such that effects are consumer's goals for a particular task when trying to
likely to persist in the same way that perceptual illusions ascertain why his or her choice processes take a certain
persist (Tversky and Kahneman 1981, 1988). Thus, in- form.
centives may be less effective for problems involving A fruitful level of analysis for developing an explana-
perceptual factors than for problems involving accuracy- tory framework is to examine consumers' metagoals for
effort trade-offs. Finally, Simonson and Tversky (1992) choice processing. Examples of such metagoals are max-
explicitly invoke the analogy to perceptualcontrasteffects imizing the accuracy of a decision, minimizing the cogni-
in motivating their trade-off contrast results. tive effort requiredfor the decision, minimizing the expe-
Although the accuracy-effort and perceptual frame- rience of negative emotion while making the decision, or
works, considered separately, can each account for some maximizing the ease with which a decision can be justi-
findings in constructive choice, we believe that an inte- fied. Note, however, that the usefulness of a choice goal
grated framework that extends each approach and then framework is compromised if too many goals are postu-
combines the two approaches is both possible and would lated, such as a different goal for each decision. To gain
be extremely useful. The key to integrating the ap- explanatorypower, we focus on the limited subset of such
proaches is to note that the perceptual approachhas much goals postulated above and examine the degree to which
to say about which aspects of a choice task are noticed such goals can explain observed constructive decision
and how tasks are represented, and the accuracy-effort making.
approach is well suited for considering how consumers The selection of these particulargoals is not arbitrary;
utilize the information they notice in order to attain their we believe that as a set they capture many of the most
goals. Thus, the two approaches complement each other importantmotivational aspects relevant to decision mak-
ing. Historically (e.g., in the rational choice approach), Attention, Information Selectivity, and Perceptual In-
accuracywas the firstgoal considered; the goal of making terpretation. As noted earlier, the fact that consumers
a choice was considered to be maximizing utility. Simon have limited processing capacity means that they gener-
(1955) and others were instrumental in bringing effort- ally cannot process all of the available information in a
relatedgoals to bear on understandingchoice in response particular situation. Hence, selectivity is necessary, and
to the realization that humans have limited processing which information is selected for processing can have a
capacity,as discussed above. Computationalproblems are major impact on choice. Put another way, it is critical to
not the only factors that make choices difficult, however. understand the determinants of the focus of attention,
Humans are emotional beings, and choices can involve since many contingent choice effects are brought about
wrenching trade-offs. Thus, we believe that the goal of by making salient different aspects of the choice environ-
minimizing experienced negative emotion is importantin ment.
some situations. Finally, humans are also social beings, Based on the psychology of attention, we know that
and one of the most decision-relevant characteristics of there are two major types of attention, voluntary and in-
the social context is that decisions are often evaluated, voluntary (Kahneman 1973). Voluntary attention de-
either by others or by oneself. Hence, the decision maker scribes when attention is devoted to information that is
often must be able to justify a decision to others or to perceived to be relevant to current goals (i.e., is labeled
himself or herself. as "diagnostic" in Feldman and Lynch's [1988] accessi-
We argue that different subsets of these goals are rele- bility-diagnosticity framework). Individuals will devote
vant in different situations. We will specify further some more effort to examining information they believe will
of the many problem characteristicsthat influence a deci- help them attain whichever goals are more heavily
sion when various goals are relevant. For example, irre- weighted in that situation.
versible decision problems that are very importantto one- Attention also may be capturedinvoluntarilyby aspects
self or one's spouse may evoke goals for increased of the environment that are surprising,novel, unexpected,
accuracy, minimizing negative emotion, and increased potentially threatening, or extremely perceptually salient,
ease of justification. The relative weight placed on various thus exemplifying one aspect of accessibility in the Feld-
goals will also reflect the individual's ability to obtain man and Lynch (1988) framework.For example, changes
timely and unambiguous feedback on his or her perfor- and losses may be particularly salient (Kahneman and
mance relative to these goals. In general, effort feedback Tversky 1979), and particular problem representations
is much easier to obtain than accuracy feedback. The may make certain aspects stand out and gain involuntary
individual usually has a very good notion about how hard attention.Thus, attentionand selectivity can be influenced
he or she is working and thus has timely and unambiguous both by goal-driven and more involuntaryperceptual fac-
feedback on effort, whereas feedback on accuracy is often tors (for a similar distinction, see Tversky [1977]).
delayed and ambiguous (Einhorn 1980). The consumer The effects of goals and perceptual factors can go be-
also can assess his or her emotional state fairly easily. yond attention. For example, preexisting goals can lead
Finally, feedback on ease of justification is often immedi- to motivated reasoning and distortion of the meaning of
ate but can be ambiguous, since what makes for a good new information (Kunda 1990; Russo, Medvec, and
explanation in a given situation may not be clear. Reason- Meloy 1996). In addition, aspects of the environment
able justifications may be easier to predict in situations that capture involuntary attention may also set in motion
where the choice task is relatively simple with limited perceptual interpretationsand behavioral responses (e.g.,
information,because the options for justification are more interpretinga loud noise as a threat and the corresponding
limited. Thus, the ability to obtain timely and relatively orienting response). As we discuss further below, we
unambiguous feedback on effort and emotion may make believe that particularlysalient aspects of the choice task
those goals particularly salient in many situations. How- can at times not only capture attention but also suggest
ever, there has not been nearly enough research on the certain types of heuristics. For example, simple problem
factors determiningthe relative salience of goals in choice displays that make asymmetric dominance visually trans-
situations. We summarize with two propositions: parent may make heuristics that are based on relational
Proposition1.1: Consumersmake choices in orderto properties of the options more accessible and thus more
accomplishgoals. Four of the most importantgoals for likely to be used. We believe that the connection of both
consumerdecisionmakingare (a) maximizingthe accu- goal-driven and more perceptual processes to selectivity
racy of the choice, (b) minimizingthe cognitive effort and interpretationis a critical factor enabling a more inte-
requiredto makethechoice,(c) minimizingtheexperience
of negativeemotionwhenmakingthechoice,and(d) max- grated framework for understanding constructive con-
imizingthe ease of justifyingthe decision. sumer choice. To summarize:
Proposition1.2: The relativeweight placed on various
goals will be influencedby a varietyof problemcharacter- Proposition1.3:Two majorinfluenceson the selectivity
istics, includingthe importanceand irreversibilityof the of attentionare currentgoals (voluntaryattention)and
decisionandthe timelinessandambiguityof the feedback surprising,novel, threatening,unexpected,or otherwise
availableon performancerelativeto each goal. perceptuallysalientaspectsof the choiceenvironment(in-
voluntary attention). Such factors can also affect the per- Proposition 1.4: Individuals have a repertoire of differ-
ceptual interpretationof focal aspects of the environment. ent strategies for solving decision problems. This repertoire
will vary across individuals depending on their experience
Choice Heuristics. First, we assume that individuals and training.
Proposition 1.5: Different strategies vary in their advan-
have a repertoireof strategies for solving decision prob-
tages and disadvantages with respect to accomplishing dif-
lems, perhaps acquired through experience or training. ferent goals in a given situation. These relative advantages
Different consumers may vary in terms of the strategies and disadvantages may be a function of consumers' skills
they possess; for example, many children's processing and knowledge and will vary from one choice environment
deficits may be due to lack of knowledge of appropriate to another.
strategies(John and Whitney 1986; Roedder 1981; Roed- Proposition 1.6: Consumers select the strategy that best
der, Sternthal, and Calder 1983). meets their goals for a particular situation given the array
Second, different strategies vary in their advantages of possible strategies and their advantages/disadvantages
and disadvantageswith respect to accomplishing different with respect to these goals.
goals in a given situation. That is, different strategies will This general outline of our choice goals approach
be more or less accurate, effortful, emotionally shows how constructive processing may occur. However,
wrenching, or easy to justify in a given choice environ- to better understandthe approach,we examine how it can
ment. For example, the weighted adding strategy may be made more specific. In particular, first we examine
tend to be accurate, effortful, and potentially more emo- situations where maximizing accuracy and minimizing
tionally difficult because it requires making trade-offs, effort are the two major focal goals and show how our
which may be emotion laden in some situations. It is less approach can explain such decisions. Then we consider
clear how weighted adding would fare in terms of ease situations where the goal of minimizing negative emotion
of justification; its thorough processing would aid justifi- is also relevant. Finally, we examine a particularlyinter-
cation, but the many subjective trade-offs required could esting case where both perceptual factors and the goal of
hinder justification. Elimination-by-aspects, on the other ease of justification are relevant.
hand, may be easy to explain and defend (Tversky 1972),
avoids emotion-laden trade-offs, and varies in its effort
and accuracy depending on characteristics of the choice Analyzing Situations Where Accuracy
task. and Effort Goals Predominate
Third, the advantages and disadvantages for a given For many consumer choices, there is little emotional
strategy will be affected by individual differences in com- involvement or need to justify. In such situations, we and
putational skills and expertise in the choice domain, both others argue that two preeminent goals for a decision are
of which can affect how easily and accurately a heuristic maximizing the accuracy of the decision and minimizing
can be implemented (Stanovich and West 1998). For the cognitive effort involved in reaching that decision
example, the ability to analyze and select the most rele- (e.g., Beach and Mitchell 1978; Hogarth 1987; Payne,
vant information improves with expertise (Alba and Bettman, and Johnson 1993; Shugan 1980).
Hutchinson 1987; Russo and LeClerc 1994; West, Brown,
and Hoch 1996); with increased expertise a consumer Assessing Cognitive Effort and Accuracy. The choice
might be more able to evaluate the safety of a car, for goals framework is most useful if we have conceptually
instance. appropriateand easily calculable measures of the various
Fourth, the relative advantages and disadvantages for goals and the extent to which different strategies accom-
a particularstrategy may differ from one environment to plish these goals in different task environments. For ex-
another. For example, a strategy that is more accurate ample, we have proposed measures of cognitive effort and
in one environment may be less accurate in another, or accuracy. With respect to cognitive effort, any decision
different information presentationformats may make cer- strategy can be decomposed into more elementary infor-
tain strategies more or less effortful to implement (for a mation processes (EIPs), such as reading an item of infor-
classic example of such format effects, see Russo mation, comparing two items of information, multiplying
[1977]). or adding items of information, eliminating items of infor-
Finally, given the array of possible approaches and mation, and so on (see Chase [1978] for a general discus-
their relative advantages and disadvantages for a given sion of using EIPs to analyze information processing). A
situation, we assume that the consumer selects the ap- lexicographic strategy, for example, could be conceptual-
proach that best meets his or her goals for that situation. ized as reading the value for each attributeweight, com-
Thus, consumers may select different approaches in dif- paring the weight just read with the largest weight found
ferent situations as their goals, the constraintsof the situa- previously until the most important attribute has been
tion, and/or their knowledge change (for a discussion of found, and then reading the values for the options on that
the potential infinite regress problem of "deciding how attribute and comparing them until the largest value is
to decide how to decide . . . " see Payne et al. [1993], found. A weighted adding strategy could be thought of
pp. 107-108). In summary, we propose: as reading weights and values, multiplying the two, mov-
CONSTRUCTIVE CONSUMER CHOICE PROCESSES 195
ing on to the next weight and value and multiplying them, FIGURE 1
adding the products, and so on. Thus, for any given strat- SELECTION OF STRATEGIES WITH CONSTRAINTS
egy, a representationin terms of EIPs can be developed. ON EFFORT
When the given strategy is applied to make a selection
WADD
in a given situation, how many EIPs of each type were 1.0 O -_
required to make that selection can be determined. The
number of EIPs used will be a function of the specific -i~,EQV,\A Preference Function
rule, the size and other characteristicsof the problem, and
the specific values of the data (see Payne et al. [1993], ?j .75 -
chap. 3, for detailed examples). Cognitive effort can then
< ! LEX
be defined as a function of the number and types of EIPs
needed to complete a task (i.e., some EIPs require more
effort than others). This approachto measuring cognitive >> .5
effort has been validatedby showing that the time to make X EBA
a decision using a specified strategy and individuals' self-
reports of the effort required are modeled quite well by I .25
weighted counts of the EIPs characterizing that strategy
(Bettman, Johnson, and Payne 1990). a) EffortConstraint
The accuracy of a decision strategy can be defined by
using the weighted adding model, which is a normative 0 E:?RC
model in that it specifies how individuals can best reflect 200 150 100 50 0
their preferences if certain assumptions about those pref- Effort(Total ElPs)
erences are met (Keeney and Raiffa 1976). A decision
process that selects the same option as the weighted add- NOTE.-WADD = weighted adding; LEX = lexicographic; EBA = elimina-
tion-by-aspects; EQW - equal weight; RC = random choice; ElPs = elemen-
ing rule is therefore an accurate process. Accuracy can tary information processes.
also be defined by avoidance of choice patterns such as
intransitivities or selection of dominated options. Finally,
accuracy can be defined by examining properties of the stance, computationalerrorsare more likely for strategies
decision process. Choice processes such as weighted add- like weighted adding than for the lexicographic heuristic;
ing that are both compensatory and extensive, utilizing errors in keeping one's place in the process may be more
all relevant information, are typically considered more likely for strategies such as EBA, where the consumer
normative (Frisch and Clemen 1994); see Hammond must keep track of exactly which options have already
(1996) for a discussion of alternative conceptions of been eliminated. Such computational errors and memory
choice accuracy. Given these measures of cognitive effort or "bookkeeping" operations have generally not been
and accuracy, we can characterizethe accuracy and effort included in the models to date. In addition, computational
of various strategies in particular environments by run-
difficulties due to the complexity of the ratings for the
ning a computer simulation of each strategy in each envi- options (e.g., ratings that involve unwieldy fractions)
ronment and keeping track of the number of EIPs used have not been modeled, although there is evidence that
for each choice and the option chosen (Johnson and Payne such complexity can influence processing (Johnson,
1985). Payne, and Bettman 1988). Finally, implementation dif-
Although a good deal of progress has been made in ficulty also depends on the individual's processing skills,
assessing the accuracy and cognitive effort of alternative which will vary as a function of such factors as socioeco-
strategies, several unresolved issues remain. For example, nomic status (Capon and Burke 1980), age (e.g., children
some strategies have not been specified at a process level
[Capon and Kuhn 1980; Gregan-Paxton and John 1995;
in terms of EIPs. An importantexample of this is Tversky
Roedder 1981] or the elderly [Cole and Balasubramanian
and Simonson's (1993) componential context model and
1993]), and expertise (Alba and Hutchinson 1987).
its potential variants discussed above. In general, rela-
Therefore, modeling implementation difficulty is another
tional heuristics that might arise in choice problems that
major area for research in specifying the accuracy and
are more perceptual in nature have not been formulated
effort of strategies more precisely.
in process terms. One goal for future research, therefore,
is to develop process versions of such heuristics in terms Using Effort and Accuracy Values to Make Predictions.
of EIPs (e.g., the componential context model would in- To show how such values might then be used, consider
volve reads, difference operations, comparisons, addi- the hypothetical example shown in Figure 1. This figure
tions, and so on). Then the effort and accuracy required shows the average accuracy (measured in terms of the
for such heuristics could be simulated. percentage attainedof the optimal weighted adding value)
Also, work to date has not adequately taken the diffi- and average effort (in total EIPs) for five strategies:
culty of implementing strategies into account. For in- weighted adding, lexicographic, EBA, equal weight, and
randomchoice. A preference function (reflecting the deci- FIGURE 2

sion maker's relative emphasis on the goals of accuracy EFFORT AND ACCURACY LEVELS FOR VARIOUS
and effort) has also been shown; although it has been STRATEGIES FOR DIFFERENT DECISION ENVIRONMENTS
depicted as linear for the sake of simplicity, a nonlinear
monotonic function could be used as well. Finally, we WADD
have also indicated an effort constraint. 1.0 O
For the moment, ignore the effort constraint. With the
EQVPLEX Preference Functions
given preference function, a consumer would select the
equal weight strategyfor this environment.We can model
the effect of changes in the decision maker's relative ci .75
0 750
EQW
emphasis on accuracy versus effort by changing the slope
of the preference function. For example, as the preference ?LEX
function became less steep (reflecting a greater relative > .5 0 EBAk
desire for accuracy), the consumer would eventually se- c EBA
lect weighted adding; if it were made more steep (re-
flecting a greater relative concern for effort), the con-
sumer could eventually select EBA or even random a) .25 Decision EnvironmentA Decision EnvironmentB
EU 0
choice. Research supportssuch predictions;decision mak-
ers instructedto maximize decision accuracy shift to more
extensive, less selective, and more alternative-basedpro- 0 RC
cessing when compared to decision makers instructed to 200 150 100 50 0
minimize effort (Creyer, Bettman, and Payne 1990;
Effort (Total ElPs)
Payne, Bettman, and Luce 1996).
Now consider the original preference function and the NOTE.-WADD= weighted adding;LEX= lexicographic;EBA= elimina-
effort constraint. For that constraint, only the lexico- tion-by-aspects;EQW= equalweight;RC= randomchoice; ElPs = elemen-
taryinformationprocesses.
graphic, EBA, and random choice strategies are feasible,
and the lexicographic strategy would be selected. Thus,
this simple example shows how variables such as time
pressure might influence strategy selection by imposing and relative weights on accuracy and effort goals [Creyer
an effort constraint. et al. 1990]). In each case the processing results were in
Finally, we consider how choice strategies might shift line with the predictions derived by using accuracy and
depending on characteristicsof the decision environment. effort goals to predict strategy usage. We will examine
In Figure 2 we plot a second set of accuracy and effort in more detail how this approach may apply to specific
values for the decision strategies in a second decision consumer decision-making situations when we review the
environment. This pattern of values emphasizes that the findings on constructive consumer choice. Next, however,
accuracy and effort characterizing a given strategy can we examine consumer choice situations where negative
be highly dependent on the situation. In this case the emotion may be evoked and consider the implications of
lexicographic strategy has become more accurate and the adding the goal of minimizing negative emotion to those
equal weight strategy less accurate in environment B. for accuracy and effort.
Such changes might occur if the variance across the
weights characterizing the attributes for the options was Analyzing Situations Where Minimizing
high in environment B and low in environment A. Thus,
the heuristic best meeting the consumer's accuracy and
Negative Emotion Is Relevant
effort goals can vary across environments. For the exam- Consumers sometimes face emotion-laden choices.
ple in Figure 2, equal weight would be selected for envi- Such choices arise when there are choice conflicts be-
ronment A, but the lexicographic strategy would be used tween goals that are very importantto the individual (e.g.,
for environment B. Note that although we focus on spe- one cannot attain all goals given the set of available op-
cific strategies in this discussion, the results also provide tions and must give up something on one importantgoal
some insights into the kinds of processing to be expected to attain more of another important goal). In such cases,
(e.g., the prevalence of attribute-based or alternative- trade-offs are required that the individual does not want
based processing, selectivity, and extent of processing) to make, since trade-offs in such situations involve giving
if consumers selected strategies according to this frame- up attainmentof some goal on which the individual does
work. not wish to accept a loss. Examples of such emotion-
We have carried out the approach outlined above to laden consumer choices include trading off the safety of
examine the effects of several different features of deci- an automobile against environmental concerns (if larger
sion environments (e.g., time pressure [Payne et al. 1988, vehicles fare better in crashes but worse in gas mileage)
1996], interattributecorrelation [Bettman et al. 1993], or trading off health risks due to the presence of insects
in one's house versus health risks from having chemicals the possibility that one attributemust be sacrificed to gain
sprayed in one's yard. Such choices can easily lead to on another, but alternative-based processing highlights
negative emotion, since the trade-offs required represent such trade-offs.
threats to the attainment of important or valued goals Our choice goals approach argues that in negatively
(Lazarus 1991). The degree of emotion often depends emotion-laden choices, consumers will simultaneously
on the values of the options (e.g., the degree of conflict process more extensively (reflecting the goal of accuracy)
and which specific attributesare involved in the conflict). and in a more attribute-basedfashion (reflecting a goal
Note that the nature of emotion-laden choices is such of coping with negative emotion), a pattern that in fact
that the ensuing negative emotion is associated with the has been supported (Luce, Bettman, and Payne 1997).
decision itself and not with some unrelatedambient affect The conceptual analyses above and these results support
such as negative mood attributableto background noise the notion that examining how other importantgoals, such
at the site where the decision must be made (see Isen as minimizing negative emotion, interact with goals for
[1997] for a discussion of affect and decision making). maximizing accuracy and minimizing effort can lead to
We believe that individuals may cope with emotion- insights into consumer decision making (Garbarino and
laden decisions by altering processing to escape negative Edell [1997] demonstrate another way in which such
emotion. Choice processes under negative emotion may goals may interact, arguing that expending cognitive ef-
therefore be affected by accuracy and effort concerns as fort on evaluating options can lead to negative affect and
modified by emotion minimization concerns. In particu- influence choice).
lar, two general coping strategies may apply in emotion- Note that one difference between our analysis of deci-
laden situations: problem-focused coping (direct actions sions involving emotion and our analysis of the effects
aimed at improving the person-environment relationship of accuracy and effort is that we have to date no easy
eliciting the emotion) and emotion-focused coping (indi- measure for the amount of emotion characterizinga deci-
rect actions aimed at minimizing emotion through sion. Unlike the case of measuring cognitive effort ad-
changes in the amount or content of thought about the dressed above, the amount of emotion generated in mak-
situation; Folkman and Lazarus 1988). These two forms ing a decision is not likely to be primarily a function of
of coping are typically used simultaneously (Folkman the processes involved in a strategy. Rather the degree of
and Lazarus 1988; Terry 1994). emotion will depend in a complex fashion on the content
Problem-focused coping involves direct efforts to solve of the decision (i.e., the specific attributes involved and
the problem at hand. In decision making, we argue that their properties), characteristics of the consumer (since
problem-focused coping will involve attempting to iden- what is emotion laden for one person may not be for
tify the most accurate decision alternative. This motiva- another), properties of the decision task such as the
tion to perform accurately should be particularly associ- amount of conflict, and the type of processing carriedout.
ated with extensive processing. Such extensive processing Thus, measuring the extent of emotion on-line (i.e., dur-
is the most readily available (to oneself) and observable ing a decision process) and modeling the determinantsof
(to others) indicator of one's motivation to be accurate. the emotional difficulty of a choice task are important
When asked to list factors associated with accurate deci- topics for future research on the choice goals framework
sions, individuals most often list consideration of all rele- (see Richins [1997] for work on measuring the emotions
vant information (Payne et al. 1988, p. 551, n. 4). Hence, characterizingconsumption experiences). In the next sec-
we expect that increased negative emotion due to the tion we add a fourth major goal, maximizing ease of
choice situation will lead to more extensive processing. justification, to our consideration of constructive con-
Emotion-focused coping often involves avoidant be- sumer choice.
haviors. Thus, one way in which emotion-focused coping
may be brought to bear on emotion-laden choices is Analyzing Situations Where Maximizing Ease
avoidance of those aspects of the decision that are particu- of Justification Is Relevant
larly emotion provoking. The aspect of emotion-laden
choices that is most taxing is making the difficult trade- Consumer decisions are often evaluated, either by oth-
offs required, because trade-offs call attention to losses. ers to whom one is accountable or by oneself. Hence,
Many researchershave argued that trade-offs are uncom- consumers often must be able to justify or provide reasons
fortable and are avoided when possible (Hogarth 1987; for a decision (Shafir, Simonson, and Tversky 1993).
Tetlock 1992; Tversky and Shafir 1992), and we believe Accountability and the need to justify decisions to others
this tendency is exacerbated when choices are emotion can have important effects on consumer decisions. Tet-
laden. Therefore, individuals may cope with emotion- lock (1992) has proposed a contingency approach that
laden decisions by avoiding explicit trade-offs; such allows for both effort-minimizing and accuracy-maximiz-
avoidance can be accomplished by using noncompensa- ing responses to increased accountability. In particular,
tory strategies (Hogarth 1987), particularly those non- Tetlock argues that people will cope with accountability
compensatory strategies in which processing is attribute in some situations by simply deferring to the preferences
based. Attribute-basedprocessing minimizes confronting of the person(s) to whom they are accountable, an effort-
TABLE 3 relational heuristics (such as the componential context

AN EXAMPLEOF AN ASYMMETRICDOMINANCE TASK model or its variants) are both easy to apply and provide
good reasons. The fact that option B dominates option C
Brand Ride quality Miles per gallon is a good reason for choosing option B at an outcome
level of explanation (it is clearly a better outcome than
A 83 24 C). There is no need to refer to a process-level explana-
B 73 33 tion (e.g., I chose B over A because of the trade-offs I
C 70 33
prefer between ride quality and miles per gallon). Out-
SOURCE.-This problem was taken from Simonson (1989, p. 164). comes are likely to be more salient than process in the
wake of a decision, so argumentsbased on outcomes may
provide better reasons. In fact, serious concerns often
minimizing strategy. However, people's inferences about arise in decision analysis and managerialdecision making
the preferences of others may not be very accurate. For because of this focus on outcomes and avoidance of pro-
example, Hoch (1988) has shown that consumers have cess explanations (e.g., Baron and Hershey 1988).
low accuracy when predicting the interests and opinions Thus, relational heuristics appear to perform well both
of the typical American consumer. If the preferences of with respect to ease of justification and effort goals. Even
those to whom one is accountable are not known, Tetlock though information on the relationships between options
proposes that individuals may process more thoroughly. must be generated, such relationship information is very
Even in this case, more thorough processing may lead to easy to assess and process for simple problems.Using such
poor results if the extra effort is not devoted to diagnostic relationshipinformation as a reason allows the consumer
information. to avoid making trade-offs, leading to the prediction that
We believe that the effects of a goal of maximizing asymmetricdominance effects might increase as trade-offs
ease of justification cannot be fully accounted for by accu- become more emotionally difficult (since using the asym-
racy and effort considerations, however. In particular,the metric dominancerelationshipto avoid trade-offswould be
need to consider other factors beyond accuracy and effort even more appealing). Luce (1998) reportssome empirical
is especially likely to be true for choice problems that are support for this prediction. Thus, relational notions may
more perceptual in nature (i.e., small numbers of alterna- also help minimize negative emotions in some cases.
tives and attributes depicted either graphically or using For this case of asymmetric dominance, therefore, a
simple tables of numerical ratings). Consider, for exam- complex interaction of multiple goals may explain the
ple, the choice problem among cars shown in Table 3 effects observed. Noting relationships among the options
(Simonson 1989). This choice problem is an example of and using these relationships as a basis for choice accom-
a problem involving asymmetric dominance (i.e., brand plishes some combination of minimizing effort, minimiz-
C is dominated by brand B but not by brand A), and it ing negative emotion, and maximizing ease of justifica-
is very easy to detect these relationships among the option. It is less clear how using such relationships affects
tions at a glance. In problems of this sort, adding brand accuracy; this probably depends on the exact structureof
C to a choice set originally consisting of brand A and the choice sets, suggesting that manipulating choice-set
brand B increases the choice share of the dominating structureso that the relationshipinformationeither is con-
alternative, brand B (Huber et al. 1982). If consumers sistent or inconsistent with accuracy goals is a promising
are asked to justify their choices, the strengthof this effect research direction. We suspect that such a complex inter-
increases (Simonson 1989). As noted by Tversky (1996, play of goals will be the norm for choices involving justi-
p. 17), such violations of context independence indicate fication issues.
that "people do not maximize a precomputed preference The above argumentsdepend heavily on the notion that
order, but construct their choices in the light of available asymmetric dominance relationships are generally quite
options." How can this result be explained by our choice simple and perceptual in nature and that the dominance
goals approach? relations are very easy to assess. We also argue that the
We believe that Simonson's (1989) original insight is very salience of such relations is part of what makes them
correct: that increased accountability leads consumers to good reasons. However, for problems of any complexity,
weight ease of justification more heavily, that this goal such relationshipsmay be more difficult to assess. For ex-
leads to a search for good reasons to use as justifications, ample, as the number of attributesincreases, the number
and that the asymmetric dominance relationship provides of required pairwise attributecomparisons increases, and
a good reason (anyone can easily see B is better than C). the consumer may not be able to make the comparisonsin
We now analyze further how such choice based on rea- a simple, perceptual fashion (i.e., it may not be easy to
sons (Shafir et al. 1993) fits within our framework. This take in the options "at a glance"). In these more complex
analysis is more speculative than those above, since most situations, the amount of effort expended or the goodness
research in this area has not been based on a choice goals of the outcome may be the easiest aspects of the choice for
approach. the individualor observerto assess, unlike the obviousness
We assert that for simple, more perceptual problems, and salience of relational propertiesin simpler situations.
Thus, we hypothesize that relationalreasons will be used correlation,completeness of information,informationfor-

less as problemsbecome more complex and/or those rela- mat, and comparable versus noncomparable choice. In
tions are made more difficult and less transparent.Further general, decisions become more difficult as the amount
research on how problem complexity affects the use of of information increases, as the time resources available
relationalnotions as reasons is needed. for processing the available information decrease, as the
The search for reasons may not always lead to effort degree of conflict among attributes increases, as the
savings, however. The argumentsfor effort savings above amount of missing information increases, as the informa-
focused on the use of relational properties. Developing tion display format becomes less organized or more com-
justifications may actually require more effort in other plex, and as the number of shared attributes describing
situations.For example, Luce (1998) shows that consum- the options is smaller. For each topic we first consider
ers choosing the status quo option, which has been argued the basic findings and then address how our framework
to lead to betterjustifications, actually took more time to accounts for these findings.
make a choice than consumers who did not select the
status quo. Thus, more research is needed on measuring Problem Size. Studies of the effects of problem size
the justifiability of different strategies across situations show that increases in the number of alternatives facing
and the effort those strategies require. the consumer lead to greater use of noncompensatory
We have now presented the basic ideas underlying our strategies that eliminate alternatives (e.g., Johnson and
choice goals approach.Although some aspects have been Meyer 1984; Payne 1976). Compensatory strategies,
the subject of a good deal of research (e.g., accuracy and however, are more common for small numbers of options.
effort goals), the role of negative emotion has seen more Changes in the number of attributesdo not appearto lead
limited research, and the role of ease of justification is to strategy changes as readily but may increase selectivity
admittedly more speculative. We now review the evi- (Olshavsky 1979; Payne 1976). Gregan-Paxtonand John
dence for constructive consumer choice and address how (1997b) have recently extended this line of research to
those results can be accounted for within our integrated examine decision processing by children; they find that
framework.We organize the review as we did the discus- older (ages 10- 11) children adapt by searching less ex-
sion of the framework,first considering research on situa- haustively for larger problems, but younger (ages 7-8)
tions where accuracy and effort goals appearto predomi- children adapt to larger problems only when search costs
nate, then choices where negative emotions play a major are explicitly made salient. Finally, Otnes, Lowrey, and
role, and finally addressing choices involving justification Shrum (1997), using in-depth interviews, found that
issues. In each subsection we summarize the major find- brides-to-be often (but not always) respond to the com-
ings as propositions; these propositions are summarized plexities of planning a wedding by simplification.
in Appendix B. Our framework accounts for these findings by exam-
ining the way changes in the number of alternatives
EVIDENCE FOR CONSTRUCTIVE and attributes affect the accuracy and effort for various
CONSUMER CHOICE: CHOICE TASKS strategies. Payne et al. (1993, pp. 133-137) report sim-
WHERE ACCURACY AND EFFORT ulation results showing that the relative accuracy of
GOALS PREDOMINATE heuristics is fairly robust over changes in the number
of alternatives, that the relative accuracy of heuristics
Many findings in research on consumer decision mak- other than weighted adding is sensitive to the number
ing can be accounted for in terms of accuracy and effort of attributes, and that the effort required to implement
goals. For instance, one naturalfocus, given an emphasis other heuristics increases much less rapidly with prob-
on accuracy and effort goals, has been on factors influ- lem size than the effort required for weighted adding.
encing problem difficulty and how consumers cope with Thus, heuristics other than weighted adding will appear
such difficulty. In additionto effects of problem difficulty, relatively more efficient as the number of alternatives
researchers have also examined one other important as- increases; in addition, selective, attribute-based heuris-
pect of the decision task itself: the type of response re- tics such as lexicographic and EBA also appear rela-
quired of consumers (response mode). Recall that ratio- tively attractive under these conditions (Payne et al.
nal choice theory assumes that preferences will not vary 1993). These simulation results are consistent with the
across strategically equivalent methods of eliciting prefer- empirical results reported above.
ences. Finally, there has been research addressing how A critical issue related to problem size is whether con-
the knowledge individual consumers bring to decision sumers can be given too much information. The issue
tasks affects decision making. of possible information overload has been of continued
interest in consumer research, with major information
Consumer Choice Research provision implications for both marketersand policy mak-
on Problem Difficulty ers. For example, if providing more information could
We consider six main subtopics under research on lead to harmful effects, then policies requiring new infor-
problem difficulty: problem size, time pressure, attribute mation might have impacts opposite to those intended.
200 JOURNALOF CONSUMERRESEARCH
Initial research on information load (Jacoby, Speller, and graphic or EBA; when time pressure is severe, it appears
Kohn 1974a, 1974b) claimed that consumers could be that quickly examining at least some information on each
overloaded and that consumers made poorer decisions option is more effective than examining a limited set of
with more information. Several researchers rapidly de- options in depth. Similar findings are reported by Payne
bated this conclusion and argued that consumers in those et al. (1996) when time stress was manipulatedby vary-
studies in fact made more accuratedecisions if the amount ing the opportunitycost of delaying decisions; by Pieters,
of informationper alternativewas increased (Russo 1974; Warlop, and Hartog (1997) for visual scans of brand
Summers 1974; Wilkie 1974); accuracy only suffered displays; and by Eisenhardt (1989) for firms processing
when more alternatives were added. Others have subse- alternatives in high-velocity environments. Some fasci-
quently found decreases in accuracy as the amount of nating research supportingsuch breadth-rather-than-depth
information per alternative was increased as well (Keller strategies is provided by Jacoby et al. ( 1994). Jacoby and
and Staelin 1987; Malhotra 1982), but controversy still his colleagues allowed consumers to access information
surroundsthe interpretationof these results (Keller and about birth control options and measured consumers' de-
Staelin 1989; Meyer and Johnson 1989). A major prob- gree of certainty regarding their choice of birth control
lem is identifying a good decision. Most studies use sub- method after each item of information was acquired. In
jects' attributeimportance ratings to ascertain the "best" one condition, Jacoby et al. controlled the format of acqui-
alternative;however, these ratings are subject to error,and sition to be either alternative based or attribute based.
this error often covaries with the amount of information Consumers' feelings of uncertainty were reduced more
(Meyer and Johnson 1989). during within-attribute searches; when search was con-
Given these problems in defining accurate choices, we strained to be within-alternative, uncertainty remained
suggest a reorientation of research on information load. high until almost all options were examined. Since all
Rather than focus on accuracy directly, we believe that options cannot be examined in depth in many instances
the essence of consumer response to information load of time pressure, these results provide corroboratingevi-
is selectivity. The critical issue, then, is how consumers dence for consumers' preferences for within-attribute
become selective as load increases. That is, how do prob- search under time pressure.
lem size and other factors (e.g., expertise, information Such shifts to attribute-basedprocessing can be pre-
format, perceptual factors) interact to influence consum- dicted based on accuracy and effort concerns. Effort is
ers' decisions about which informationto select for exam- essentially fixed under time pressure. Payne et al. (1988)
ination? If consumers become selective in ways that re- also show that accuracy decreases markedly for weighted
flect their values, then overload may not be too harmful. adding under severe time pressure because there is not
However, if selectivity is based on surface features of the enough time to complete processing, yet the accuracy of
decision task such as perceptual salience or format, and attribute-basedheuristics such as lexicographic and EBA
these surface features are not reflective of consumers' is much more robust under time pressure.
underlying values, then substantial decreases in accuracy Time pressure has other effects on shifting the focus
are possible. In summary: of attention. For example, Wright (1974) reports that
consumers place greater emphasis on negative informa-
Proposition 2.1: Increasesin the numberof alternatives
leadto a greateruse of noncompensatory choice strategies. tion about options under time pressure. This shift in focus
Increasesin the numberof attributesgenerallyleads to of attention could be due to both goal-driven (e.g.,
increasedselectivity,but not strategychanges. avoiding bad options) or involuntary (e.g., threats being
Proposition 2.2: Information loadwill not be harmfulto very salient) factors. The major results for time pressure
accuracyif consumersselect a subsetof the information can be summarized as:
thatreflectstheirvalues;if selectivityis basedon problem
Proposition 2.3: Undermoderatetimepressure,consum-
featuresthatdo not reflectconsumers'underlyingvalues, ersprocessmorerapidlyandbecomemoreselective.Under
informationload can decreaseaccuracy. moreseveretimepressure,individualsswitchto moreattri-
bute-basedheuristics.
Time Pressure. As the pace of life accelerates, con- Proposition 2.4: Consumersweightnegativeinformation
sumers may feel that they are subject to increased time moreheavilyundertime pressure.
pressure. Time pressure has clear effects on choice pro-
cesses. Payne et al. (1988) found that individuals engaged Attribute Correlation. Interattributecorrelation is re-
in a hierarchy of responses when faced with increased lated to notions of dominance; when the number of attri-
time pressure. Under moderate time pressure, individuals butes is small, removing dominated options leads to more
processed each item of informationmore rapidly (acceler- negative interattributecorrelation for the remaining op-
ation) and became somewhat selective. Under more se- tions (Curry and Faulds 1986). Interattributecorrelation
vere time pressure, people accelerated their processing, is also related to the concept of conflict among the attri-
became more selective, and changed their decision strate- bute criteria:the more negative the correlations, the more
gies. In particular,they switched to more attribute-based one has to give up something on one attributein order to
processing, which characterizes such strategies as lexico- get more of another attribute.
Bettman et al. (1993) show that many simplifying heu- sumers used other-brandinformation,namely, the average
ristics become less accurate, compared to the weighted value of the attributefor the other options, discounted to
adding model, when interattribute correlations become take account of the fact that there was uncertainty. John-
more negative. Thus, if accuracy concerns are salient, our son and Levin (1985) arguedthat same-brandinformation
framework predicts that consumers may shift to more was used, and Ford and Smith (1987) also found that
extensive, less selective, and more alternative-basedpro- consumers were influenced more by information about
cessing under negative correlation. other attributes of the same brand than by other-brand
Bettman et al. (1993) found this hypothesized shift information.
toward more normative processing under negative corre- Otherresearch,however, has taken a contingency view-
lation, but replication in a consumer context is needed. point, arguing that whether other-brand or same-brand
Widing and Talarzyk (1993) provide suggestive evidence information is used is a function of properties of the
supportive of the hypothesized relationship between use choice task. Several researchers have invoked the Feld-
of the weighted adding strategy and correlation. They man and Lynch (1988) accessibility-diagnosticity frame-
showed that a decision aid that provided weighted average work as the basis for their contingency approaches. Dick,
scores using importanceweights provided by the user was Chakravarti,and Biehal (1990), for example, argue that
highly favorably evaluated by users and resulted in the consumers for whom attributeinformation is more acces-
best decision accuracy in a negative correlation environ- sible are more likely to make inferences using interattri-
ment (better than a decision aid based on cutoffs and an bute correlation (same-brand information). Ross and
unaided format). In contrast, Johnson, Meyer, and Ghose Creyer (1992) propose that consumers use other-brand
(1989) found that their subjects did not appear to shift information when it is diagnostic, perhaps because it is
strategies; however, they also manipulatedthe number of less effortful, only examining same-brandinformation if
options, and subjects may have focused on that manipula- other-brandinformationis not diagnostic. However, some
tion and may have failed to perceive correlation accu- researchershave sounded cautionarynotes; Simmons and
rately (Huber and Klein 1991; Klein and Yadav 1989). Lynch (1991), for example, note that the discounting
Although there is some evidence that negative correla- effects noted above may not be mediated by inferencing.
tion may be associated with more alternative-basedpro- Broniarczyk and Alba (1994) point out that consumers'
cessing, this relationship may not hold for all types of intuitive beliefs about likely attribute relationships may
decision tasks. Luce, et al. (1997, experiment 3) showed be dominant in inferencing, even when they conflict with
that processing became more attribute based as correla- the available same-brand or other-brandinformation.
tions became more negative when the decisions were The application of the choice goals framework to these
more emotion laden. We speculate that negative correla- results is relatively straightforward.Consumers have heu-
tion leads to more alternative-based processing for less ristics for making inferences about missing information
emotional decision tasks but to more attribute-basedpro- (e.g., same-brandor other-brand). The use of such heuris-
cessing for more emotion-laden decisions, because alter- tics appears to be governed by accessibility-diagnosticity
native-based processing is likely to elicit negative emo- trade-offs. Feldman and Lynch (1988) define the diagnos-
tion by highlighting difficult trade-offs in the latter case. ticity of an input in terms of the degree to which the
To summarize: decision implied by that input would accomplish the indi-
Proposition2.5: Consumersengagein processingmore vidual's goals for the choice in question (e.g., maximize
like weighted adding (more extensive, less selective, and accuracy, maximize ease of justification). Effort is also
more alternativebased) under negative interattributecorre- a major component of their approach, both with respect
lation in less emotional decision tasks. to accessibility and the termination of search by a diag-
nosticity threshold. Since accuracy has been a salient goal
in many accessibility-diagnosticity studies, accessibility-
Completeness of Information. Consumers may wish diagnosticity trade-offs may be essentially effort-accuracy
to have a complete set of information about the brands trade-offs in those situations. However, in other situations
and attributes they choose to consider. However, this is accessibility-diagnosticity concerns can also refer to a
often not the case. What happens when one or more op- balance among the broader set of choice goals we have
tions are partially described (i.e., there is missing infor- proposed. Finally, the Broniarczyk and Alba (1994) re-
mation)? Decision makers might respond in a variety of sults are explainable by noting that generalizing from
ways to such a circumstance. The consumer might try to intuitive beliefs is even less effortful than using same-
infer the missing value based on other available informa- brand or other-brandinformation.
tion. Two such types of available informationfor inferring
a missing attribute value for a given option are other Proposition 2.6: Consumersmayinfervaluesfor a miss-
values for that attributein the set of options (other-brand ing attributeusingheuristicsbasedon other-brand informa-
information) and values for other attributesof the given tion (other values for that attributein the set of options)
option (same-brand information). Early work (e.g., or same-brandinformation(values for otherattributesof
Huber and McCann 1982; Meyer 1981) argued that con- the same brandwith the missing value). Which type of
heuristic is used depends on the relative accessibility and (1994), for example, shows that consumers will restruc-
diagnosticity of each type of information. ture information if it enables them to facilitate choice.
Restructuringis more likely when information is poorly
InformationFormat. The organizationof information structured and hard to process in its original form. The
displays can have a major impact on consumer choices. specific form of restructuringdepends on the sources of
Russo (1977) provides a classic illustration; he argued difficulty (e.g., consumers may rescale information if it
that normal unit price displays, with separatetags for each is presented in different units). Thus, the form in which
item, were difficult to process when attempting to make consumers represent the choice problem is a function of
price comparisons. He argued that making such compari- accuracy and effort considerations.
sons easier by providing unit price information in the The consumers' task goals may also affect where the
form of a sorted list, with brands ranked in order of in- concreteness principle holds. Sethuraman,Cole, and Jain
creasing unit price, would increase consumers' usage of (1994) show that concreteness holds if the consumer's
unit prices. In a field experiment, he showed that consum- task is to screen a set of options to see which are accept-
ers saved about 2 percent, on average, with the list dis- able, since this task can be done without regard to order
play. Thus, consumer decision making was affected by of presentation of information. However, if the task is
making some information easier to process. Simply mak- to choose the first acceptable option (satisficing), then
ing information easier to process is not always sufficient, information is processed by brand regardless of the for-
however. Russo et al. (1986) found no effects from listing mat. Thus, accuracy considerations can override the po-
positive nutrients (e.g., vitamins), but did find effects if tential effort savings. Finally, Biehal and Chakravarti
negative nutrients (e.g., sugar) were listed. To use the (1982) show that memory organization can interact with
lists consumers must be motivated, even if they are easily format; prior brand-based memory organization, for ex-
processable; consumers apparently did not feel they ample, weakens the effects of attribute-basedformats.
needed to worry about positive nutrients (they could take These results can be readily explained within our
a vitamin pill) but were motivated to avoid negative ones. framework by noting that different information formats
Finally, Bettman, Payne, and Staelin (1986) outline prin- make some forms of processing easier and less effortful
ciples for providing warning information in less effortful and hence more likely. The concreteness principle and
formats that encourage processing of such warnings. governance of the processing of information by the form
A related display effect concerns the impact of display of the display are also explainable by cognitive effort
format on information acquisition. Slovic (1972) sug- considerations when the other acquisition costs and poten-
gested a "concreteness" principle: decision makers will tial accuracy losses are low. However, if processing is
tend to use only that information that is explicitly dis- difficult in the original format or there are potential accu-
played and will use it in the form it is displayed, without racy losses, consumers may process in ways that depart
transforming it. Such behavior reduces the cognitive ef- from the format. Summarizing these results, we propose
fort requiredto process the information.A related concept
is the so-called focusing phenomenon in mental models, Proposition2.7: Differentinformationformatscanmake
in which thoughts are restrictedto what is explicitly repre- some forms of processingeasier and less effortfulthan
others.The processingencouragedby the formatwill be
sented in the mental model (Legrenzi, Girotto, and John- morelikely;however,potentialaccuracylosses due to the
son-Laird 1993). In addition, when individuals construct encouragedform of processingwill lessen the effect of
mental models of a situation, they often make as little format.
information as possible explicit in order to minimize cog-
nitive load.
The concreteness principle is supported by Bettman Comparable versus Noncomparable Choices. For
and Kakkar's (1977) finding that individuals indeed ac- many years consumer choice researchersfocused on deci-
quired information in a manner consistent with the form sion processes in choice sets where the alternatives were
of the display (by attributeor by brand), and Jarvenpaa members of the same product class, such as selecting
(1989) extended these results to the case of graphical among brands of automobiles or microwaves. Johnson
displays. A similar finding is that sequential presentation (1984, 1986) pioneered work on choice among noncom-
of informationleads to more significant pioneering effects parable options. For example, a consumer might try to
(e.g., more learning about the pioneer and greater likeli- decide whether to spend some extra money on a vacation,
hood of including the pioneer in the consideration set; a new stereo, or some new cooking equipment. Such op-
Kardes and Kalyanaram 1992; Kardes et al. 1993). tions are called noncomparablebecause the attributesthat
Several authors show that the concreteness effect does describe them differ across the options. Ratneshwar,
not always hold, however. Concreteness is likely to hold Pechmann, and Shocker (1996) examined conditions that
when the costs of following the given format (both the might lead to such noncomparablechoice sets and found
acquisition costs and opportunity costs due to losses in that consumers developed consideration sets with alterna-
accuracy) are low. If those costs are more pronounced, tives from different product categories when there was
consumers may depart from the given format. Coupey goal conflict (one category could not satisfy all of the
consumer's goals) or goal ambiguity (no clear, salient Kahneman 1990). A major contributing factor to prefer-
goals). ence reversals seems to be that some response modes,
Processing can differ between comparable and non- such as choice, evoke qualitative (e.g., comparative) rea-
comparable choices. Johnson (1984) showed that as op- soning, whereas other response modes, such as evalua-
tions become more noncomparable, consumers represent tion, evoke more quantitative (e.g., how much more will
attributesat more abstract levels (e.g., enjoyment or ne- I pay) reasoning (Fischer and Hawkins 1993; Slovic,
cessity) in order to allow comparisons. At some point, Griffin, and Tversky 1990). Nowlis and Simonson
however, consumers switch to developing overall evalua- (1997) provide supporting evidence for the latter notion.
tions of each option and comparing those evaluations. They show that attributeson which consumers can easily
Bettman and Sujan (1987) argued that a fundamental make comparisons (comparable attributes) assume more
distinction between comparable and noncomparable weight in choice, whereas attributes that are difficult to
choice situations is that criteriaare more readily available compare but are meaningful and informative when evalu-
for the comparablechoice. When a criterion (e.g., reliabil- ated on their own (enriched attributes) take on more
ity) was primed for the noncomparable situations, deci- weight in tasks where individual options are evaluated.
sion processes resembled those for the case of comparable This difference in emphasis across response modes leads
options. In related work, Huffman and Houston (1993) to systematic preference reversals. Coupey, Irwin, and
showed that the presence of goals helps to structureand Payne (1998) have recently shown that an individual's
focus the acquisition of information, even in unfamiliar familiarity with the stimuli used can affect the frequency
situations, and West, Brown, and Hoch (1996) found that of preference reversals; in particular, consumers make
providing consumers with a consumption vocabulary led fewer choice-matching preference reversals for more fa-
to better-definedpreferences. miliar products. The biggest impact of familiarity in their
These results can be explained by invoking effort and study was on consumers' responses to the matching task.2
accuracy notions. In fact, Johnson (1986) developed As the above discussion implies, judgment and choice
models for the effort and accuracy of the within-attribute are not equivalent. When making a judgment, consumers
abstraction and across-attributeoverall evaluation strate- must provide an overall evaluation of each alternative;in
gies and derived conditions under which each strategy choice, the consumer must simply pick the most preferred
was likely to be used based on accuracy-efforttrade-offs. option. A judgment task thus encourages less selective,
The Bettman and Sujan (1987) results can be explained alternative-based processing, whereas a choice task
by noting that their priming task makes a particularcrite- allows more selectivity and can be attribute based, de-
rion more accessible, hence enabling consumers to choose pending on the strategy used.
based on that concrete criterion as opposed to making Such response mode effects are difficult to undo. In a
more effortful inferences about more abstract attributes. classic set of studies, Gretherand Plott (1979), for exam-
ple, found that monetaryincentives did not eliminate pref-
Proposition 2.8: Processing can differ between compara-
erence reversals. In general, incentives based on maximiz-
ble and noncomparable choices. Consumers can develop
more abstract attributes or compare overall evaluations as ing accuracy often have relatively little impact on
choices become more noncomparable. Their use of these decisions (Tversky and Kahneman 1988), perhaps be-
strategies is governed by accuracy-effort concerns. If a cause incentives appearto lead individuals to increase the
specific criterion is primed, however, noncomparable intensity of their behavior (i.e., working harder) without
choice processes more closely resemble those for compara- necessarily changing the direction of that behavior (i.e.,
ble options. working smarter). However, providing incentives and
feedback over repeated choices does appear to reduce
preference reversals (Cox and Grether 1996).
Consumer Choice Research The analysis of response mode effects within our
on Response Mode framework is more speculative than the analyses pre-
sented above. We argue that response mode alters choice
A fundamental principle of rational decision theory is because it both changes the criterion used for task accu-
procedure invariance, that is, that preferences should not racy and the relative emphasis on accuracy and effort.
depend on the method used to assess them if the methods The response mode determines the consumer's criterion
are strategically equivalent (Tversky et al. 1988). How- for the task. That is, as noted above, choice requires de-
ever, this principle does not seem to hold empirically; termining which option is better, whereas other response
different methods can strongly influence expressed prefer- modes such as matching and ratings require a quantitative
ences. Mowen and Gentry (1980) show, for example, that estimate of how much better one option is over another
preferences between two new product projects could be or how much to give up on one attribute for more of
reversed depending on whether the decision maker was
asked to choose one of the projects or designate a price
for the rights to each project. Such preference reversals 2In a matching response, an aspect of one option is adjusted so that
are common in many contexts (e.g., Tversky, Slovic, and this option matches another option in overall value.
another. We believe that this difference in focus means asked to recall the information. Consumers who first
that there is more relative emphasis on accuracy when learned the information tended to organize it in memory
matching or rating and more relative emphasis on effort by brand and were more likely to process information in
when making a choice. Since we have shown above that an alternative-basedfashion when making choices. Con-
changes in the relative emphasis on accuracy versus effort sumers who chose first were more likely to process by
can change strategy usage, it is perhaps not surprising attribute and to organize their memory in an attribute-
that preference assessments might differ across response based fashion. In addition, consumers had much better
modes.3 memory about the chosen alternative (see also Johnson
We can speculate further, however. We hypothesize and Russo 1984). In their 1983 paper, Biehal and Chakra-
that there will be fewer preference reversals between varti again had consumers either learn information or
choice and some quantitative task (e.g., rating or match- make a choice. Then information on new alternatives and
ing) to the degree that the use of heuristics emphasizing new attributes was provided and a second choice made.
accuracy is encouraged in the choice task. There is sug- The added attributeinformation made one of the original
gestive evidence consistent with this hypothesis. Johnson options, which was not likely to be selected originally,
et al. (1988) found that preference reversals were lower more attractive.Consumers who made a choice first were
when processing of the probability information in their less likely to reevaluate and choose that original option,
task was made easier by expressing the probabilities as perhaps because of their decreased memory for the non-
simple decimals or simple fractions instead of as complex chosen option, than the consumers who learned first.
fractions. Coupey et al.'s (1998) results showing a de- These studies provide convincing evidence that memory
crease in preference reversals with familiarity also seem and choice processes interact in interesting and principled
supportive. One corollary of our hypothesis is that if a ways (see also Lynch and Srull 1982).
matching task evokes a strategy with greater weight on Other research has continued to examine memory and
one attribute, the results of matching may agree more choice interactions. Lynch, Marnorstein, and Weigold
with the choice task, since choice is more likely to be (1988) showed that consumers will use information in
lexicographic in nature. Hence we hypothesize that there memory to make a choice when some or all of the options
will be fewer choice-matching preference reversals if the had to be retrieved from memory if that information is
match is carried out on the more important versus the both accessible and more diagnostic than other accessible
less important attribute. information. In related work, Nedungadi (1990) showed
that brandchoice can be influencedby alteringthe memory
Proposition 3.1: Differentresponsemodes (e.g., choice accessibilityof a brand,withoutchangingthe brand'sevalu-
vs. matching)canleadto differentpreferencesandto pref- ation. Hutchinson, Raman, and Mantrala (1994) model
erence reversals.Choice encouragesmore selective and
attribute-basedcomparativeprocesses,whereasratingor such brand name accessibility as a function of consumer
matchingencourageless selective,morealternative-based, and marketingmix variables (e.g., usage rate, advertising
quantitativeprocessing. spending, marketpenetration,and product attributes).
Proposition 3.2: Preferencereversalscan be reducedby We explain such choice-memory interactionsby invok-
makingthe morequantitativeprocessingeasier,by greater ing effort considerations. For example, Biehal and Chak-
familiaritywith the options,and by repeatedexperience ravarti's learning tasks encourage an alternative-basedfo-
with incentivesandfeedbackon the task. cus, and alternative-basedprocessing would then be less
effortful in choice. Choice, in contrast, tends to be more
comparative and attribute based. Thus, consumers who
Consumer Research on the Interactions chose first would find it more difficult to integrate the
between Consumer Knowledge new and old information, because their choice processing
and Decision Making was both more selective and more attribute based. For
consumers that learned first, however, the information
Choice Process and Memory Interactions. Biehal would be stored and processed more by alternative, so
and Chakravarti(1982, 1983) carried out a classic series the added information would be easier to integrate with
of studies examining the interactionsbetween choice pro- each option. Finally, work based on accessibility-diagnos-
cesses and memory. Biehal and Chakravarti(1982) had ticity notions, as noted above, can be viewed in terms of
consumers either learn information and then choose or balancing choice goals.
just choose; following the choice task, consumers were Proposition 4.1: Becausechoice processesare selective
and comparative,choice processesinfluenceboth which
items are storedin memoryand the organizationof such
3Cognitive effort considerations continue to play a role in judgment,
memories.Conversely,informationin memorycan affect
even though there may be more relative emphasis on accuracy. There the formof choice processes.
is a great deal of evidence for compatibility effects in judgment, i.e., Proposition 4.2: Consumerswill use informationin
that the weight of an attributein a decision is enhanced by its compatibil- memoryto make a choice to the extent that it is more
ity with the response scale or strategy used (Fischer and Hawkins 1993; accessibleandmorediagnosticthanalternativesourcesof
Slovic et al. 1990). information.
Categorization and Choice. Fiske (1982; Fiske and do not always lead to accuracy; for example, the use of
Pavelchak 1986) has arguedthatindividuals use two basic analogies may lead to insensitivity to the quantity of a
modes for evaluating stimuli. In one mode, piecemeal good (Baron and Greene 1996). Characterizingthe prop-
processing, a stimulus is evaluated by integratingthe eval- erties of analogies in terms of effort, accuracy, experi-
uations of the individual attributes of the stimulus. In enced negative emotion, and ease of justification, and
contrast, category-based processing is characterized by then examining the contingent use of analogies would be
attemptingto categorize the stimulus into an existing cate- promising areas for future research.
gory; if successful, that category's evaluation is associ-
ated with the stimulus. Fiske and Pavelchak propose a Proposition 4.3: Categorical processing is more likely
to be used when expectations are met, with more detailed
two-stage contingency approach:in the first stage, catego-
(piecemeal) processing more likely when expectations are
rization is attempted and, if successful, then category- not met and the consumer has expertise in the category.
based evaluation is used. If categorization is not success-
ful, piecemeal processing ensues. Sujan (1985) utilized This concludes our review of research where accuracy
this approach in a consumer context, providing evidence and effort goals predominate. Next we consider research
for category-based processing when the information in a where the goal of minimizing negative emotion must be
print advertisement matched expectations and evidence taken into account.
for piecemeal processing when expectations were not met,
with effects greater for experts than for novices (see
Gregan-Paxton and John [1997a] for a modification and EVIDENCE FOR CONSTRUCTIVE
extension of this approach based on internal knowledge CONSUMER CHOICE: CHOICE TASKS
transfermechanisms). Peracchio and Tybout (1996) also WHERE MINIMIZING NEGATIVE
examined the effects of categorization on evaluations. EMOTION IS RELEVANT
They demonstrated that new products were evaluated
more favorably when their attributeswere moderately in- As discussed above in presenting our framework, we
congruent with a schema (see also Meyers-Levy and Ty- believe that at times consumers face emotion-laden
bout 1989) only when consumers had limited knowledge choices. We have examined the effects of such negative
about the product category. When consumers had more emotion on decision processing, choice, and avoidance of
extensive knowledge about the category, their evaluations choice. Luce et al. ( 1997) showed that increased negative
were influenced by associations to specific attributes emotion due to more difficult trade-offs or more serious
based on their schemas but not by the degree of congruity. consequences led to more extensive, more selective, and
Finally, John and Lakshmi-Ratan(1992) apply categori- more attribute-basedprocessing (consistent with the no-
zation notions to children's choice. tion that consumers try to avoid emotion-laden difficult
The categorizationfindings can be explained by consid- trade-offs). Luce, Payne, and Bettman (1998) showed
ering the accuracy and effort characteristics of the two that consumers more often chose an option that was better
classes of strategies, categorization and piecemeal. Cate- on a quality attributeratherthan an option better on price
gorization is less effortful and relatively accurate when when that quality attributewas rated as inherently more
expectations are met and categorization is successful. emotion laden, involved lower attribute values, or in-
Piecemeal processing is more effortful and generally ac- volved losses rather than gains. These effects were in
curate. Hence, if categorization can be successfully ap- addition to any effects of the consumers' importance
plied, it is both less effortful and relatively accurate, so weights for quality relative to price. Luce (1998) showed
it will be used. Sujan's (1985) finding that novices are that increases in negative emotion led to increased avoid-
more likely to use categorical processing regardless of ance of choice; in particular,increased trade-off difficulty
whether expectations are met can also be explained. The led to increased choice of a status quo option, increased
cost of piecemeal processing is less for experts than for choice of an asymmetrically dominating alternative, and
novices because of the experts' increased ability; hence, a greater tendency to prolong search. In addition, she
experts are more likely to use piecemeal processing when demonstrated that choosing an avoidant option resulted
expectations are not met. in less negative retrospective emotion. Finally, Kahn and
Gregan-Paxtonand John (1997a) have arguedthat con- Baron (1995) found that consumers were likely to use a
sumers make use of analogies. We also believe that analo- noncompensatory, lexicographic rule when making their
gies can be used as a choice heuristic. For example, own high-stakes decisions, but that consumers wanted
Schkade and Payne (1994) have shown that consumers advisors to make decisions for them using compensatory
use analogies such as charitable giving when expressing rules.
their preferences for environmental goods. The use of Such results have been explained above in describing
analogies clearly involves accuracy and effort considera- the part of our framework dealing with negative emotion.
tions; an analogy can provide a way of making a choice However, there are two importantpoints still to be made.
in an unfamiliar situation with little effort, with accuracy First, the results show that individuals process in a more
depending on the goodness of the analogy. Analogies extensive, more selective, and more attribute-basedfash-
ion when choices are more emotion laden. Note that this Proposition 5.1: Emotion-ladenchoices are character-
pattern of more extensive and more attribute-basedpro- ized by moreextensive,moreselective,andmoreattribute-
cessing is opposite to that found in all of our studies of based processing.In general,emotion-ladenchoices en-
nonemotional choices, where more extensive processing courageavoidantbehaviors.
is linked to more alternative-basedprocessing. Recall also
that negative correlationappears to have different effects
on processing for less emotional than more emotion-laden EVIDENCE FOR CONSTRUCTIVE
decisions. Taken together, these differences in results be- CONSUMER CHOICE: CHOICE TASKS
tween less and more emotion-laden decision tasks demon- WHERE MAXIMIZING EASE OF
stratethat negative emotion due to the decision task can- JUSTIFICATION IS RELEVANT
not be treated as simply another type of effort cost.
Emotional decision costs appear to have an additional As mentioned earlier, consumers are often not sure how
motivational component that leads to avoidant behaviors. to trade off one attribute relative to another or which
Second, we have argued that trade-off difficulty plays attributeis more importantwhen confronted with a deci-
a crucial role in the degree to which a choice is viewed sion. Shafir et al. (1993) have argued that people often
as emotion laden and can affect processing and choice; construct reasons in order to resolve their conflict about
however, we have not yet examined the nature of trade- trade-offs, justify a decision to themselves (i.e., increase
off difficulty. The traditional view of trade-offs is that their confidence in the decision), and/or justify (explain)
"a trade-off was a trade-off was a trade-off" (Tetlock, their decision to others. Such a reasons-based view of
Peterson, and Lerner 1996, p. 36). This point of view choice is clearly consistent with the view of constructive
assumes that people are willing and able to make trade- preferences. In addition, the justification of decisions is
offs among any conflicting attributes and that the only one of the important choice goals in our framework.
relevant question is the exchange rate between the attri- Shafir et al. (1993) argue that relationships among op-
butes. However, Tetlock and his colleagues, along with tions may be perceived to be more compelling reasons or
others like Baron (1986; Baron and Spranca 1997) and argumentsfor choice than deriving overall values for each
Beattie and Barlas (1993), have argued that the identity option and choosing the option with the best value. For
of the conflicting attributes is in fact very relevant and example, Montgomery (1983) has argued that a domi-
that people will find some trade-offs much more difficult nance relationship between options is easy to justify as a
than others. Tetlock (1992; Tetlock et al. 1996) gives reason for choice. He suggests that as a consequence
examples of such difficult attributes to trade off using decision makers may restructureproblems (e.g., discount
what he terms "sacred" values (e.g., life, justice, or lib- small attributedifferences or drop attributesof lesser im-
erty). People resist putting prices on life or justice, for portance) in order to achieve a dominance relationship.
example. Baron and Spranca (1997) discuss "protected" Other relationships that may provide good reasons are the
attributes,which are attributes that resist trade-offs with best value on the most important attribute or principles
other attributes. of rational choice such as transitivity.
There is some limited evidence in the consumer be- Of particularinterest to consumer research is the fact
havior literature that certain attributes are treated dif- that changes in the set of options under consideration
ferently than others. In particular, consumers generally change the relationships among the options and therefore
seem to be more resistant to trading off some quality some of the potential reasons for choosing among the
to get a better price than accepting a higher price to get options. That is, changes in the decision context can alter
better quality (Blattberg and Wisniewski 1989; Hardie, the reasons that are salient and, consequently, the choice
Johnson, and Fader 1993; Heath and Chatterjee 1995; made. Research on asymmetric dominance is an excellent
Nowlis and Simonson 1997; Simonson and Tversky exemplar of such context-dependentpreferences based on
1992). Dhar and Simonson (1997) also distinguish be- relationships among options.
tween situations involving trade-offs between two goals One of the classic assumptions of models of choice is
(e.g., pleasure and health) and those involving trade- the independence of irrelevant alteruatives assumption.
offs between goals and resources (e.g., pleasure and The basic idea behind this assumption is that the ratio of
money) and show that choices vary depending on the the choice probabilities for any pair of options does not
type of trade-off involved. change if the composition of the choice set containing
Although some research has been done on attribute the two is changed. As noted by Meyer and Kahn (1991,
properties, we believe that much more research is needed. p. 91), the "central implication is that if a new option is
We believe that attributesdiffer in fundamentalways with added to a choice set, the shares of the existing options
respect to trade-offs (e.g., some attributesare more sacred will always decrease in direct proportion to the size of
or more laden with potentially severe consequences) and their original shares." A weaker condition than the inde-
that research characterizing attribute properties and the pendence of irrelevant alternatives (and hence one for
effects of those properties on trade-off difficulty, pro- which violations are a more serious problem) is regularity,
cessing, and choice is extremely important. the notion that adding a new alternative cannot increase
the probability of choosing a member of the original aversion reflects that intermediate options tend to be fa-
choice set. vored because disadvantages loom larger than advantages
It is clear that consumers' choices do not always obey and intermediate options have relatively smaller disad-
such assumptions; the composition of a choice set (the vantages than extreme options.
context) does affect choice behavior. As noted above, We have attemptedto explain such results above when
for example, asymmetric dominance has the remarkable presenting our framework. Our basic argumentis that for
effect that adding an asymmetrically dominated option to simple choice problems the relations among the choice
the choice set actually increases the choice share of the options are both easy to ascertain perceptually and are
dominating alternative, in violation of the principle of likely to be viewed as plausible reasons or justifications
regularitynecessary for most probabilistic choice models. for choice. Indeed, we might push the argument a step
Huber et al. (1982) originally demonstrated this effect, further and assert that situations with simple options
and Heath and Chatterjee (1995), Huber and Puto where the relationships among the options are easy to
(1983), Mishra, Umesh, and Stem (1993), Simonson ascertain are more likely to evoke ease of justification as
(1989), and Simonson and Tversky (1992) have pro- a relevant choice goal. Because they can be perceived
vided additional demonstrations. with little effort, relationships among options may be used
Several alternative explanations have been proposed to make choices in situations with simple options even if
for the asymmetric dominance effect, including agenda they do not provide good justifications (e.g., for compro-
effects (Huber et al. 1982), choice simplification by mise effects, see Simonson [1989]).
searching for dominant options (Montgomery 1983), and As noted above, we believe that relationships among
the simplicity and artificialityof the stimuli (Ratneshwar, options will be more difficult to assess if the set of choice
Shocker, and Stewart 1987; however, see Wedell [1991] options is large or displayed in such a way that these
for evidence contradicting this view). Prelec, Wernerfelt, relationships are less transparent.We also believe that the
and Zettelmeyer (1997; see also Wemerfelt 1995) argue value component of the componential context model is
that at least part of the effect is due to the fact that con- more relevant to attaining accuracy goals and that the
sumers develop inferences about the location of their ideal relative advantagecomponent is more relevant for achiev-
points based on the stimuli and their relationships to one ing effort and justification goals. Recall that 0 is the pa-
another. rameter in the componential context model that is the
We believe that many causes contributeto asymmetric weight given to the relative advantage component; that
dominance effects. However, one explanation that has re- is, 0 may capture the degree to which the focus is on
ceived strong support is that people use the relationships local relationships in a choice set rather than absolute
among the options as reasons for choice justification:one criteria. Based on the discussion above, we hypothesize
can justify the choice of the dominatingoption more easily that 0 will be a function of the relative strength of goals
by noting that it is clearly better than the asymmetrically (with 0 higher if less weight is placed on accuracy and
dominatedoption (Simonson 1989; Wedell 1991). Recall more weight is placed on effort and justification), task
that Simonson (1989) found that increasedneed for justifi- size (with 0 lower the greater the number of options and
cation led to a greaterasymmetricdominance effect. attributes),4 and transparencyof the relationships (with
As described above, Tversky and Simonson (1993; Si- 0 higher the more transparentthe relationships). We be-
monson and Tversky 1992) developed a componential lieve that research testing such predictions regarding the
context model accounting for asymmetric dominance and componential context model would be very important.In
related context effects with two components: a measure summary, we assert:
of the value of each alternative if considered separately Proposition 6.1: Adding an asymmetrically dominated
and a measure of the relative advantage of each option alternative to a choice set results in increased choice share
compared to all of the other options in the choice set. for the dominating alternative. This effect is enhanced un-
This lattercomponent explains effects such as asymmetric der increased need for justification.
dominance, since the dominating option has a clear ad- Proposition 6.2: Many context effects can be accounted
vantage over the asymmetrically dominated option, for by considering the value of each alternative considered
whereas the advantages of the nondominated alternative separately and the relative advantage of each option com-
over the dominating and asymmetrically dominated op- pared to the other available options.
tion are not as clear.
Simonson and Tversky (1992), in a tour de force paper, DISCUSSION
demonstrate that the principles of trade-off contrast and
We have now presented a framework for constructive
extremeness aversion can be used in their model to ex-
consumer choice and have attemptedto review the litera-
plain such context phenomena as local and background
contrast and attraction,compromise, and polarization ef-
fects. Trade-off contrast refers to the fact that a given 4Note the parallel to the results in noncomparablechoice, where con-
trade-off can appearmore or less favorable depending on sumers switch to relying more on overall value as the options become
otherimplied trade-offs in the choice set, and extremeness more noncomparable.
ture on consumer decision making in light of that frame- should be extended to address the principles governing
work. However, there are several issues that still remain. how representationsare formed. For example, individuals
First, there are some areas of research that we feel we may construct representations that minimize cognitive
cannot yet do a good job of explaining within our frame- load, perhaps by using relevant analogies. Depending on
work. Second, some broad generalizations about the liter- the goodness of the analogy, this can be an efficient solu-
ature have become apparent that point to large gaps in tion, although individuals may not consider alternative
our knowledge. Third, we discuss the conditions under representationsonce they have one that seems reasonable
which constructive choices are likely to be adaptive (i.e., (Legrenzi et al. 1993). When no appropriate analogy
intelligent, if not optimal, responses to task demands). exists, the consumer must construct a representation.We
Finally, we consider the implications of a constructive speculate that the representation formed will depend on
view of consumer choice for measuring preferences, one what is most salient in the choice environment (due to
of the most crucial applied problems in consumer re- both voluntary and involuntary attention), but very little
search. is known about this process.
A second area of research that is problematic for our
Findings That Our Framework model focuses on choices that have consequences over
time. Simonson (1990), for example, shows that whether
Cannot Fully Explain one chooses items for future consumption on multiple
Two areas of research pose problems for our frame- occasions by making a combined choice all at one time
work, framing effects and preferences over time. It is or separately on each occasion has a systematic influence
clear that different ways of framing a problem can lead on choices. In particular, choices made all at one time
to different choices (Tversky and Kahneman 1988). For increase the amount of variety in the choices. Read and
example, Levin and Gaeth (1988) show that labeling beef Loewenstein (1995) provide a model and theoretical ac-
as 75 percent lean results in more favorable evaluations count of this phenomenon based on time contraction(con-
than referring to it as 25 percent fat, especially before sumers compress future time intervals when making com-
tasting it. Mellers, Schwartz, and Cooke (1998) suggest bined choices and hence overestimate the effect of
that such framing manipulations influence the salience of satiation) and choice bracketing (combined choices are
the good and the bad features of outcomes. Different framed together as a portfolio, whereas the separate
frames clearly can make gains or losses more salient, and choices are viewed more independently). As another ex-
differential reactions to gains and losses underlie many ample of choices over time, Wertenbroch (1998) argues
framing effects (e.g., Puto 1987; Shiv, Edell, and Payne that consumers exert self-control in consumption by ra-
1997; Tversky and Kahneman 1981). Thus, frames may tioning their purchase quantities (e.g., buying only a sin-
have an emotional component, and we have argued above gle pack of cigarettes at a time to help cut down on their
that avoidance of losses may help satisfy the goal of smoking; see Hoch and Loewenstein [1991] for other
minimizing negative emotion (Luce et al. 1998). More work on self-control). Again, some aspects of these re-
generally, framing effects suggest that "people choose, sults can be understood in terms of our framework,partic-
in effect, between descriptions of options ratherthan be- ularly the Wertenbrochresults. The consumer appears to
tween the options themselves" (Tversky 1996, p. 7). be trading off increased acquisition effort for increased
The frameworkpresented in this article can make some long-run accuracy. However, there are also trade-offs be-
inroads into explaining such effects, arguing that a frame tween long-run and short-runaccuracy (i.e., the consumer
makes certain aspects both more perceptually salient (in- wanting a cigarette now but knowing it is bad for him or
voluntary attention) and less effortful to process. In addi- her in the long run).
tion, people may fail to transformproblems into a single These two areas of research (framing and choice over
canonical representationthat would avoid framing effects time) reveal gaps that our framework still cannot fill. In
due to limits on intuitive computation, even in simple both cases, perceptual principles seem involved to some
problems (Kahneman and Tversky 1984). People tend to extent (e.g., diminishing sensitivity to changes and loss
accept the frame presented in a problem and evaluate aversion in framing and time contraction in the choices
options in terms of the reference point suggested by that over time). Thus, research that attempts to buttress the
frame, a concept clearly related to the concreteness princi- perceptual aspects of the choice goals framework is
ple mentioned earlier. The choice goals framework, how- needed.
ever, is silent on why losses loom larger than equivalent
gains (loss aversion). Some Generalizations from Reviewing
A topic related to research on framing that our frame-
Constructive Consumer Choice
work does not address in sufficient detail is the nature of
problem representations.Ourframeworktends to consider In reviewing the literature on constructive consumer
problem representations as given by the structure of the choice, we have drawn many generalizations about spe-
task (see Coupey [1994], and Lynch, Chakravarti,and cific topic areas, expressed as the propositions summa-
Mitra [1991] for exceptions). However, our framework rized as Appendix B. However, two broader generaliza-
tions have also emerged: (1) task differences between how choice is constructed "on the fly," analysis at this
research based on accuracy-effort ideas (e.g., task diffi- level of detail seems necessary. Second, there is little
culty issues) and research based on perceptual notions research on sequences of choices over time. Without
(e.g., asymmetric dominance and other relationships studying such sequences and the interactions among
among options), and (2) the lack of research on the dy- choices and their outcomes over time, it is difficult to
namics of constructive choice. observe how representations change or how learning
First, striking task differences became apparent be- occurs at the process level, for example. Third, people
tween research based on the accuracy-effort and per- may have limited ability to predict to-be-experienced
ceptual approaches. Research based on accuracy-effort utility (Kahneman 1994). Consequently, people may
concerns generally considers larger, more complex seek to maximize the accuracy of a choice but instead
problems and focuses more heavily on process than experience later disappointment or regret due to a fail-
on outcome. Conversely, research from the perceptual ure to consider how their tastes might change over time.
perspective relies more on small, relatively simple Hence, a second broad gap in research is studying vari-
problems and focuses on outcomes rather than process. ous facets of choice dynamics.
These differences are very understandable; for exam-
ple, to observe process one must have problems of some
complexity. Otherwise there will be either no process When Will Constructive Choice Be Adaptive?
to observe or the process will be so constrained that it One issue we have not yet addressed is when construc-
will be trivial and invariant across conditions. Also, tive choices will be adaptive. Although being adaptive is
studies interested in observing process often restrict hard to define, we generally mean making intelligent, if
information gathering so that each piece of information not necessarily optimal, choices. We believe that consum-
examined can be observed (e.g., by using computerized ers are often adaptive in the above sense but that such
information acquisition programs). At the other ex- adaptivity is not universal. There is some evidence, for
treme, perceptual studies often make all information example, that consumers' responses to time pressure are
available simultaneously. adaptive. However, many factors can lead to failures in
We believe that these differences in tasks across types adaptivity, including lack of knowledge of appropriate
of studies have major consequences. For example, the strategies, difficulties in assessing propertiesof the choice
use of simple problems and simultaneous presentation in task, overreliance on the most salient factors in the task
perceptual research makes relational aspects much more environment, overgeneralization of heuristic strategies,
salient, hence making context effects and an ease of justi- and difficulties in implementing strategies. We also be-
fication goal more prevalent. Thus, one large gap in the lieve that certain kinds of failures are more likely than
literaturecould be addressed by research that attempts to others. For example, overreliance on the most salient sur-
unconfound the conceptual approach from problem type. face properties of the task can be particularlyproblematic
For example, research on context effects could be carried if such properties capture consumers' attention but are
out with larger, more complex problems. Researchers us- not related to consumers' underlying values. In such a
ing a perceptualframeworkcould also attemptto enumer- case, choices will likely not be adaptive. As a broad gener-
ate the types of relational heuristics consumers might use alization, we believe lack of adaptivity is often more prev-
and try to observe the processes involved in implementing alent in studies based on perceptual notions, partly be-
such heuristics (this would also require more complex cause such studies enhance perceptual salience and are
problems). Accuracy-effort researchers, in contrast, often designed to specifically document biases in choice.
might attempt to analyze relational heuristic properties It is clear, however, that much more research is needed
and how their usage varies depending on typical accuracy- on the extent and determinants of consumer adaptivity.
effort variables such as time pressure, problem size, and One major problem in conducting such research will be
so on. to develop defensible measures for the goodness of a
A second broad generalization that is readily appar- consumer's choice.
ent is that there is very little work on choice dynamics.
This is true at two levels. First, there is a need for work
on how the focus of attention changes constructively Assessment of Preferences
over the course of a decision. Verbal protocols taken in a Constructive World
during choice processes reveal a complex dynamic
course, with various aspects being noticed, which then In a constructive world, preferences differ depending
cause shifts in processing, which then lead to different on a wide variety of factors, as reviewed above. What
aspects being noticed, and so on (see, for example, does this imply for the assessment of preferences,
Payne et al. 1993, pp. 175-176). However, we are not which is a major focus of market research? We believe
aware of any research in consumer choice that attempts that the answer to this question depends on the goal of
to model the detailed, changing focus of attention over the analyst. One possible goal is that the analyst wishes
the course of a choice episode. To really understand to be able to predict the consumer' s preferences in order
to be able to predict market response. Context matching intriguing research opportunities remains, and we be-
is the recommended approach for prediction purposes. lieve that understanding consumer decision processes
In implementing context matching, the analyst attempts will continue to be a major focus of consumer behavior
to determine the relevant factors that might influence research and theory.
preferences in the consumer's environment and then
match the values of those factors in the measurement
environment (Simmons, Bickart, and Lynch 1993; [Received September 1997. Revised May 1998. Robert
Wright and Kriewall 1980). For example, such factors E. Burnkrant served as editor, and Barbara Loken
as the number of options and attributes, the context served as associate editor for this article.]
provided by competing options, interattribute correla-
tions, time pressure, response mode, and others can
affect choices. The environment in which preferences
are elicited should try to approximate the consumer's
APPENDIX A
environment on all of these factors, particularly if the Summary Propositions for the Integrated
consumer has little familiarity with the decision. Con- Conceptual Framework
text matching thus demands a thorough knowledge of
the properties of consumer choice environments. In Proposition 1.1: Consumers make choices in order
some cases, factors may differ systematically across to accomplish goals. Four of the most important goals
environments (e.g., the set of available options may for consumer decision making are (a) maximizing the
vary from store to store or from region to region). In accuracy of the choice, (b) minimizing the cognitive
that case, measurements may need to be taken for each effort required to make the choice, (c) minimizing the
of the major variants of the choice environment and experience of negative emotion when making the
then aggregated based on the relative frequency of these choice, and (d) maximizing the ease of justifying the
variants. Thus, context matching seems appropriate, decision.
even if potentially difficult, for prediction, which may
be the goal in many market research applications.
Proposition 1.2: The relative weight placed on var-
However, in some situations we may wish to aid
ious goals will be influenced by a variety of problem
consumers as they construct their preferences (e.g.,
characteristics, including the importance and irrevers-
when measuring values for environmental goods or
ibility of the decision and the timeliness and ambiguity
when helping a consumer select a college). For exam-
of the feedback available on performance relative to
ple, consumers are increasingly being asked to make
each goal.
choices among or provide values for environmental
goods, such as cleaning up a lake, preventing deaths
of animals, or decreasing levels of pollution. Environ- Proposition 1.3: Two major influences on the selec-
mental goods are often unfamiliar and often involve tivity of attention are current goals (voluntary attention)
difficult trade-offs. In such situations we want to help and surprising, novel, threatening, unexpected, or other-
consumers achieve a "defensible" expression of their wise perceptually salient aspects of the choice environ-
preferences (Gregory et al. 1993) or help them develop ment (involuntary attention). Such factors can also affect
preferences by considering the implications of those the perceptual interpretationof focal aspects of the envi-
preferences and how to manage them (e.g., to reduce ronment.
regret; Simonson 1992; Slovic 1995). Thus, an im-
portant area for future research is developing guidelines Proposition 1.4: Individuals have a repertoireof dif-
for a good preference construction process and doing ferent strategies for solving decision problems. This rep-
empirical research to document that such guidelines are ertoire will vary across individuals depending on their
indeed effective. Such guidelines might include ensur- experience and training.
ing consideration of multiple viewpoints and options,
using multiple response modes (Huber et al. 1993), Proposition 1.5: Different strategies vary in their ad-
and requiring explicit trade-offs. For an example of vantages and disadvantageswith respect to accomplishing
such guidelines, see Payne, Bettman, and Schkade different goals in a given situation. These relative advan-
(1998). tages and disadvantages may be a function of consumers'
We have sought to convey the constructive nature of skills and knowledge and will vary from one choice envi-
consumer choice in this article. We have also tried to ronment to another.
communicate an integrated framework for understand-
ing such behavior. Finally, throughout this article we Proposition 1.6: Consumers select the strategy that
have attempted to identify gaps requiring future re- best meets their goals for a particularsituation given the
search. These potential research areas are summarized array of possible strategies and their advantages/disad-
in Appendix C. We are excited that a broad array of vantages with respect to these goals.
APPENDIX B Comparable versus Noncomparable Choice

Proposition 2.8: Processing can differ between com-
parable and noncomparable choices. Consumers can de-
PROPOSITIONS FOR CONSTRUCTIVE velop more abstractattributesor compare overall evalua-
CONSUMER CHOICE tions as choices become more noncomparable. Their use
of these strategies is governed by accuracy-effort con-
cerns. If a specific criterion is primed, however, noncom-
Choice Tasks Where Accuracy and parable choice processes more closely resemble those for
comparable options.
Effort Goals Predominate
Response Mode
ProblemDifficulty
Proposition 3.1: Different response modes (e.g.,
Problem Size choice vs. matching) can lead to different preferences and
to preference reversals. Choice encourages more selective
Proposition 2.1: Increases in the number of alterna- and attribute-basedcomparativeprocesses, whereas rating
tives lead to a greater use of noncompensatory choice or matching encourage less selective, more alternative-
strategies. Increases in the number of attributesgenerally based, quantitative processing.
lead to increased selectivity, but not to strategy changes.
Proposition 3.2: Preference reversals can be re-
Proposition 2.2: Informationload will not be harmful duced by making the more quantitative processing eas-
to accuracyif consumers select a subset of the information ier, by greater familiarity with the options, and by re-
that reflects their values; if selectivity is based on problem peated experience with incentives and feedback on the
features that do not reflect consumers' underlying values, task.
information load can decrease accuracy.
Interactions between Consumer Knowledge
Time Pressure and Decision Making
Proposition 2.3: Under moderate time pressure, con- Choice Process and Memory Interactions
sumers process more rapidly and become more selective. Proposition 4.1: Because choice processes are selec-
Under more severe time pressure, individuals switch to tive and comparative, choice processes influence both
more attribute-basedheuristics. which items are stored in memory and the organization
of such memories. Conversely, information in memory
Proposition 2.4: Consumers weight negative infor- can affect the form of choice processes.
mation more heavily under time pressure.
Proposition 4.2: Consumers will use information in
Attribute Correlation memory to make a choice to the extent that it is more
accessible and more diagnostic than alternative sources
Proposition 2.5: Consumers engage in processing of information.
more like weighted adding (more extensive, less selec-
tive, and more alternativebased) undernegative interattri- Categorization and Choice
bute correlation in less emotional decision tasks. Proposition 4.3: Categorical processing is more
Completeness of Information likely to be used when expectations are met, with more
detailed (piecemeal) processing more likely when expec-
Proposition 2.6: Consumers may infer values for a tations are not met and the consumer has expertise in the
missing attribute using heuristics based on other-brand category.
information (other values for that attribute in the set of
options) or same-brandinformation (values for other at- Choice Tasks Where Minimizing Negative
tributesof the same brandwith the missing value). Which Emotion Is Relevant
type of heuristic is used depends on the relative accessibil-
ity and diagnosticity of each type of information. Proposition 5.1: Emotion-laden choices are charac-
terized by more extensive, more selective, and more attri-
Information Format bute-based processing. In general, emotion-laden choices
encourage avoidant behaviors.
Proposition 2.7: Different information formats can
make some forms of processing easier and less effortful Choice Tasks Where Maximizing Ease
than others. The processing encouraged by the format
will be more likely; however, potential accuracy losses
of Justification Is Relevant
due to the encouraged form of processing will lessen the Proposition 6.1: Adding an asymmetrically domi-
effect of format. nated alternativeto a choice set results in increased choice
share for the dominating alternative. This effect is en- * Examining the use of relational heuristics as a func-
hanced under increased need for justification. tion of task complexity variables.
* Characterizingthe determinantsof focus of attention
Proposition 6.2: Many context effects can be ac- over the course of a decision.
counted for by considering the value of each alternative * Studying choice processes over time.
considered separately and the relative advantage of each * Examining the extent and determinantsof consumer
option compared to the other available options. adaptivity.
* Developing guidelines for preference construction
APPENDIX C processes and empirically documenting their effec-
tiveness.
Topics for Future Research on Constructive

REFERENCES
Consumer Choice
Alba, Joseph W. and J. Wesley Hutchinson (1987), "Dimen-
Research on the Choice Goals Framework sions of Consumer Expertise," Journal of Consumer Re-
search, 13 (March), 411-454.
* Specifying factors that determine the relative sa- , John G. Lynch, Jr., Barton Weitz, Chris Janiszewski,
lience of choice goals. Richard J. Lutz, Alan G. Sawyer, and Stacy L. Wood
* Specifying relational heuristics at a process level and (1997), "Interactive Home Shopping: Consumer,Retailer,
determining their effort-accuracy characteristics. and Manufacturer Incentives to Participate in Electronic
Marketplaces," Journal of Marketing, 61 (July), 38-53.
* Modeling the implementation difficulty of strategies.
and Howard Marmorstein (1987), " The Effects of
* Measuring the degree of emotion during a decision. Frequency Knowledge on Consumer Decision Making,"
* Specifying determinants of the emotional difficulty Journal of Consumer Research, 14 (June), 14-26.
of a choice task. Baron, Jonathan (1986), "Tradeoffs among Reasons for Ac-
* Examining how problem complexity affects the use tion," Journal for the Theory of Social Behavior, 16
of relationship information as reasons. (July), 173-195.
* Measuring the justifiability of various strategies. and Joshua Greene (1996), "Determinants of Insensi-
tivity to Quantity in Valuation of Public Goods: Contribu-
Research in Choice Tasks Where Accuracy tion, Warm Glow, Budget Constraints, Availability, and
and Effort Goals Predominate Prominence," Journal of Experimental Psychology: Ap-
plied, 2 (June), 107-125.
* Characterizinghow consumers become selective as and John C. Hershey (1988), "Outcome Bias in Deci-
a function of information load and other factors. sion Evaluation," Journal of Personality and Social Psy-
* Examining how the frequency of preference rever- chology, 54 (April), 569-579.
sals is related to the emphasis on accuracy in the and Mark D. Spranca (1997), "Protected Values,"
choice task. Organizational Behavior and Human Decision Processes,
* Characterizingthe properties of analogies in terms 70 (April), 1-16.
of choice goals and examining contingent analogy Beach, Lee Roy (1990), Image Theory: Decision Making in
Personal and Organizational Contexts, Chichester: Wiley.
usage.
and Terence R. Mitchell (1978), "A Contingency
Research in Choice Tasks Where Minimizing Model for the Selection of Decision Strategies," Academy
Negative Emotion Is Relevant of Management Review, 3 (July), 439-449.
Beattie, Jane and Sema Barlas (1993), "Predicting Perceived
* Characterizing attribute properties and their effects Differences in Tradeoff Difficulty," working paper, Labo-
on trade-off difficulty, processing, and choice. ratory of Experimental Psychology, University of Sussex,
Brighton BN1 9QG, England.
Research in Choice Tasks Where Maximizing Ease Belk, Russell W. (1988), "Possessions and the Extended Self,"
of Justification Is Relevant Journal of Consumer Research, 15 (September), 139-
168.
* Examining the weight placed on the relative advan- Bettman, James R. (1979), An Information Processing Theory
tage component of the componential context model of Consumer Choice, Reading, MA: Addison-Wesley.
as a function of relative goal strength, task size, and , Eric J. Johnson, Mary Frances Luce, and John W.
transparencyof relationships. Payne (1993), "Correlation, Conflict, and Choice," Jour-
nal of Experimental Psychology: Learning, Memory, and
Other Gaps in the Literature Cognition, 19 (July), 931-951.
, Eric J. Johnson, and John W. Payne (1990), "A Com-
a Addressing the principles governing the formation ponential Analysis of Cognitive Effort in Choice," Organi-
of representations. zational Behavior and Human Decision Processes, 45
* Using more complex problems and process tracing (February), 111-139.
in perceptual framework research. and Pradeep Kakkar (1977), "Effects of Information
PresentationFormaton Consumer InformationAcquisition Dhar, Ravi and Itamar Simonson (1997), "Consumption Con-
Strategies," Journal of Consumer Research, 3 (March), text Effects on Choice: Matching versus Balancing," work-
233-240. ing paper, School of Management, Yale University, New
and C. Whan Park (1980), "Effects of Prior Knowl- Haven, CT 06511.
edge and Experience and Phase of the Choice Process on Dick, Alan, Dipankar Chakravarti,and Gabriel Biehal (1990),
Consumer Decision Processes: A Protocol Analysis," "Memory-Based Inferences during Consumer Choice,"
Journal of ConsumerResearch, 7 (December), 234-248. Journal of Consumer Research, 17 (June), 82-93.
, John W. Payne, and Richard Staelin (1986), "Cogni- Drolet, Aimee (1997), "Generic Preferences," unpublished
tive Considerationsin Designing Effective Labels for Pre- dissertation, Graduate School of Business, Stanford Uni-
senting Risk Information," Journal of Marketing and Pub- versity, Stanford, CA 94305.
lic Policy, 5, 1-28. Einhorn, Hillel J. (1980), "Learning from Experience and Sub-
and Mita Sujan (1987), "Effects of Framing on Evalua- optimal Rules in Decision Making," in Cognitive Pro-
tion of Comparable and Noncomparable Altematives by cesses in Choice and Decision Behavior, ed. Thomas S.
Expert and Novice Consumers," Journal of ConsumerRe- Wallsten, Hillsdale, NJ: Erlbaum, 1-20.
search, 14 (September), 141-154. Eisenhardt, Kathleen M. (1989), "Making Fast Strategic Deci-
Biehal, Gabriel and Dipankar Chakravarti(1982), "Informa- sions in High Velocity Environments," Academy of Man-
tion Presentation Format and Leaming Goals as Determi- agement Journal, 32 (September), 543-576.
nants of Consumers' Memory-Retrieval and Choice Pro- Feldman, Jack M. and John G. Lynch, Jr. (1988), "Self-Gener-
cesses," Journal of ConsumerResearch, 8 (March), 431- ated Validity and Other Effects of Measurement on Belief,
441. Attitude, Intentions, and Behavior," Journal of Applied
and Dipankar Chakravarti(1983), "Information Ac- Psychology, 73 (August), 421-435.
cessibility as a Moderator of Consumer Choice," Journal Fischer, Gregory W. and Scott A. Hawkins (1993), "Strategy
of Consumer Research, 10 (June), 1-14. Compatibility, Scale Compatibility, and the Prominence
Blattberg, Robert C. and Kenneth J. Wisniewski (1989), Effect," Journal of ExperimentalPsychology: Human Per-
"Price-Induced Patterns of Competition," Marketing Sci- ception and Performance, 19 (June), 580-597.
ence, 8 (Fall), 291-309. Fiske, Susan T. (1982), "Schema-Triggered Affect: Applica-
Broniarczyk, Susan M. and Joseph W. Alba (1994), "The Role tions to Social Perception," in Affect and Cognition: The
of Consumers' Intuitions in Inference Making," Journal Seventeenth Annual Carnegie Symposium on Cognition,
of Consumer Research, 21 (December), 393-407. ed. Margaret S. Clark and Susan T. Fiske, Hillsdale, NJ:
Capon, Noel and Marian Burke (1980), "Individual, Product Erlbaum, 55-78.
Class, and Task-Related Factors in Consumer Information and Mark A. Pavelchak (1986), "Category-Based ver-
Processing," Journal of Consumer Research, 7 (Decem- sus Piecemeal-Based Affective Responses: Developments
ber), 314-326. in Schema-TriggeredAffect," in The Handbook of Motiva-
and Deanna Kuhn (1980), "A Developmental Study tion and Cognition: Foundations of Social Behavior, ed.
of Consumer Information Processing Strategies," Journal Richard M. Sorrentino and E. Tory Higgins, New York:
of Consumer Research, 7(December), 225-233. Guilford, 167-203.
Chase, William G. (1978), "Elementary Information Pro- Folkman, Susan and Richard S. Lazarus (1988), "Coping as a
cesses," in Handbook of Learning and Cognitive Pro- Mediator of Emotion," Journal of Personality and Social
cesses, Vol. 5, Human InformationProcessing, ed. William
Psychology, 54 (March), 466-475.
K. Estes, Hillsdale, NJ: Erlbaum, 19-90.
Ford, Gary T. and Ruth Ann Smith (1987), "Inferential Beliefs
Cole, CatherineA. and Siva K. Balasubramanian(1993), "Age
in Consumer Evaluations: An Assessment of Alternative
Differences in Consumers' Search for Information: Public
Processing Strategies," Journal of ConsumerResearch, 14
Policy Implications," Journal of Consumer Research, 20
(June), 157-169. (December), 363-371.
Coupey, Eloise (1994), "Restructuring: Constructive Pro- Frisch, Deborah and Robert T. Clemen (1994), "Beyond Ex-
cessing of Information Displays in Consumer Choice," pected Utility: Rethinking Behavioral Decision Research,"
Journal of Consumer Research, 21 (June), 83-99. Psychological Bulletin, 116 (July), 46-54.
, Julie R. Irwin, and John W. Payne (1998), "Product Garbarino, Ellen C. and Julie A. Edell (1997), "Cognitive
Category Familiarity and Preference Construction," Jour- Effort, Affect, and Choice," Journal of Consumer Re-
nal of Consumer Research, 24 (March), 459-468. search, 24 (September), 147-158.
Cox, James C. and David M. Grether (1996), "The Preference Gregan-Paxton, Jennifer and Deborah Roedder John (1995),
Reversal Phenomenon: Response Mode, Markets, and In- "Are Young ChildrenAdaptive Decision Makers?A Study
centives," Economic Theory, 7 (April), 381-405. of Age Differences in Information Search Behavior,"
Creyer, Elizabeth H., James R. Bettman, and John W. Payne Journal of Consumer Research, 21 (March), 567-580.
(1990), "The Impact of Accuracy and Effort Feedback and Deborah Roedder John (1997a), "Consumer
and Goals on Adaptive Decision Behavior," Journal of Learning by Analogy: A Model of Internal Knowledge
Behavioral Decision Making, 3 (January-March), 1-16. Transfer," Journal of Consumer Research, 24 (Decem-
Curry,David J. and David J. Faulds (1986), "Indexing Product ber), 266-284.
Quality: Issues, Theory, and Results," Journal of Con- and Deborah Roedder John (1997b), "The Emergence
sumer Research, 13 (June), 134-145. of Adaptive Decision Making in Children," Journal of
Dawes, Robyn M. (1979), "The Robust Beauty of Improper Consumer Research, 24 (June), 43-56.
Linear Models in Decision Making," American Psycholo- Gregory, Robin, Sarah Lichtenstein, and Paul Slovic (1993),
gist, 34 (July), 571-582. "Valuing Environmental Resources: A Constructive Ap-
proach," Journal of Risk and Uncertainty, 7 (October), Asim Ansari, and Tracy Troutman (1994), "Tracing the
177-197. Impact of Item-by-Item Information Accessing on Uncer-
Grether, David M. and Charles R. Plott (1979), "Economic tainty Reduction," Journal of Consumer Research, 21
Theory of Choice and the Preference Reversal Phenome- (September), 291-303.
non," American Economic Review, 69 (September), 623 - , Donald E. Speller, and Carol A. Kohn (1974a),
638. "Brand Choice Behavior as a Function of Information
Hammond, Kenneth R. (1996), Human Judgment and Social Load," Journal of Marketing Research, 11 (February),
Policy: Irreducible Uncertainty, Inevitable Error, Un- 63-69.
avoidable Injustice, New York: Oxford University Press. , Donald E. Speller, and Carol A. Kohn (1974b),
Hansen, Flemming (1972), Consumer Choice Behavior: A "Brand Choice Behavior as a Function of Information
Cognitive Theory, New York: Free Press. Load: Replication and Extension," Journal of Consumer
Hardie, Bruce G. S., Eric J. Johnson, and Peter S. Fader (1993), Research, 1 (June), 33-42.
"Modeling Loss Aversion and Reference Dependence Ef- Jarvenpaa,SirkkaL. (1989), "The Effect of Task Demands and
fects on Brand Choice," Marketing Science, 12 (Fall), Graphical Format on Information Processing Strategies,"
378-394. Management Science, 35 (March), 285-303.
Heath, Timothy B. and Subimal Chatterjee(1995), "Asymmet- John, Deborah Roedder and Ramnath Lakshmi-Ratan (1992),
ric Decoy Effects on Lower-Quality versus Higher-Quality "Age Differences in Children's Choice Behavior: The Im-
Brands: Meta-Analytic and Experimental Evidence," pact of Available Alternatives," Journal of Marketing Re-
Journal of ConsumerResearch, 22 (December), 268-284. search, 29 (May), 216-226.
Hoch, Stephen J. (1988), "Who Do We Know: Predicting and John C. Whitney, Jr. (1986), "The Development
the Interests and Opinions of the American Consumer," of Consumer Knowledge in Children: A Cognitive Struc-
Journal of ConsumerResearch, 15 (December), 315-324. ture Approach," Journal of Consumer Research, 12
and George F. Loewenstein (1991), "Time-Inconsis- (March), 406-417.
tent Preferences and Consumer Self-Control," Journal of Johnson, Eric J. and Robert J. Meyer (1984), "Compensatory
Consumer Research, 17 (March), 492-507. Choice Models of Noncompensatory Processes: The Effect
Hogarth, Robin M. (1987), Judgement and Choice, 2d ed., of Varying Context," Journal of Consumer Research, 11
New York: Wiley. (June), 528-541.
Howard, John A. and Jagdish N. Sheth (1969), The Theory of , Robert J. Meyer, and Sanjoy Ghose (1989), "When
Buyer Behavior, Reading, MA: Addison-Wesley. Choice Models Fail: Compensatory Representations in
Huber, Joel and Noreen M. Klein (1991), "Adapting Cutoffs Negatively CorrelatedEnvironments," Journal of Market-
to the Choice Environment:The Effects of AttributeCorre- ing Research, 26 (August), 255-270.
lation and Reliability," Journal of ConsumerResearch, 18 and John W. Payne (1985), "Effort and Accuracy in
(December), 346-357. Choice," Management Science, 31 (April), 394-414.
and John M. McCann (1982), "The Impact of Inferen- , John W. Payne, and James R. Bettman (1988), "Infor-
tial Beliefs on ProductEvaluations," Journal of Marketing mation Displays and Preference Reversals," Organiza-
Research, 19 (August), 324-333. tional Behavior and Human Decision Processes, 42 (Au-
, John W. Payne, and Christopher P. Puto (1982), gust), 1-21.
"Adding Asymmetrically Dominated Alternatives: Viola- and J. Edward Russo (1984), "Product Familiarity and
tions of Regularity and the Similarity Hypothesis," Jour- Learning New Information," Journal of Consumer Re-
nal of Consumer Research, 9 (June), 90-98. search, 11 (June), 542-550.
and ChristopherP. Puto (1983), "Market Boundaries Johnson, Michael D. (1984), "Consumer Choice Strategies
and ProductChoice: IllustratingAttractionand Substitution for Comparing Noncomparable Alternatives," Journal of
Effects," Journal of Consumer Research, 10 (June), 31 - Consumer Research, 11 (December), 741-753.
44. (1986), "Modeling Choice Strategies for Noncompara-
, Dick R. Wittink, John A. Fiedler, and Richard Miller ble Alternatives," Marketing Science, 5 (Winter), 37-54.
(1993), "The Effectiveness of Alternative Preference Elic- Johnson, Richard D. and Irwin P. Levin (1985), "More than
itation Procedures in Predicting Choice," Journal of Mar- Meets the Eye: The Effect of Missing Information on Pur-
keting Research, 30 (February), 105-114. chase Evaluations," Journal of Consumer Research, 12
Huffman, Cynthia and Michael J. Houston (1993), "Goal-Ori- (September), 169-177.
ented Experiences and the Development of Knowledge," Kahn, BarbaraE. and JonathanBaron (1995), "An Exploratory
Journal of Consumer Research, 20 (September), 190- Study of Choice Rules Favored for High-Stakes Deci-
207. sions," Journal of ConsumerPsychology, 4 (4), 305-328.
Hutchinson, J. Wesley, Kalyan Raman, and Murali K. Mantrala Kahneman, Daniel (1973), Attention and Effort, Englewood
(1994), "Finding Choice Alternatives in Memory: Proba- Cliffs, NJ: Prentice-Hall.
bility Models of Brand Name Recall," Journal of Market- (1994), "New Challenges to the Rationality Assump-
ing Research, 31 (November), 441-461. tion," Journal of Institutional and Theoretical Economics,
Isen, Alice M. (1997), "Positive Affect and Decision Making," 150 (March), 18-36.
in Research on Judgment and Decision Making: Currents, and Amos Tversky (1979), "Prospect Theory: An
Connections, and Controversies, ed. William M. Goldstein Analysis of Decision Making under Risk," Econometrica,
and Robin M. Hogarth, Cambridge:Cambridge University 47 (March), 263-291.
Press, 509-534. and Amos Tversky (1984), "Choices, Values, and
Jacoby, Jacob, James J. Jaccard, Imran Currim, Alfred Kuss, Frames," American Psychologist, 39 (April), 341-350.
Kardes, Frank R. and GurumurthyKalyanaram (1992), "Or- Malhotra,Naresh K. (1982), "InformationLoad and Consumer
der-of-EntryEffects on Consumer Memory and Judgment: Decision Making," Journal of Consumer Research, 8
An InformationIntegrationPerspective," Journal of Mar- (March), 419-430.
keting Research, 29 (August), 343-357. March, James G. (1978), "Bounded Rationality, Ambiguity,
, Gurumurthy Kalyanaram, Murali Chandrashekaran, and the Engineering of Choice," Bell Journal of Econom-
and Ronald J. Dornoff (1993), "Brand Retrieval, Consid- ics, 9 (Autumn), 587-608.
eration Set Composition, Consumer Choice, and the Pion- McFadden, Daniel (1997), "Rationality for Economists,"
eering Advantage," Journal of Consumer Research, 20 working paper, Department of Economics, University of
(June), 62-75. California, Berkeley, CA 94720.
Keeney, Ralph L. and Howard Raiffa (1976), Decisions with Mellers, Barbara A., Alan J. Schwartz, and Alan D. J. Cooke
Multiple Objectives: Preferences and Value Tradeoffs, (1998), "Judgment and Decision Making," Annual Re-
New York: Wiley. view of Psychology, 49, 447-477.
Keller, Kevin Lane and Richard Staelin (1987), "Effects of Meyer, Robert J. (1981), "A Model of Multiattribute Judg-
Quality and Quantityof Informationon Decision Effective- ments under Attribute Uncertainty and Information Con-
ness," Journal of Consumer Research, 14 (September), straint," Journal of Marketing Research, 18 (November),
200-213. 428-441.
and RichardStaelin (1989), "Assessing Biases in Mea- and Eric J. Johnson (1989), "Information Overload
suring Decision Effectiveness and Information Overload," and the Nonrobustness of Linear Models: A Comment on
Journal of Consumer Research, 15 (March), 510-512. Keller and Staelin," Journal of Consumer Research, 15
Klein, Noreen M. and Manjit S. Yadav (1989), "Context Ef- (March), 498-503.
fects on Effort and Accuracy in Choice: An Inquiry into and BarbaraE. Kahn (1991), "Probabilistic Models of
Adaptive Decision Making," Journal of Consumer Re- Consumer Choice Behavior," in Handbook of Consumer
search, 15 (March), 411-421. Behavior, ed. Thomas S. Robertson and Harold H. Kassar-
Kunda, Ziva (1990), "The Case for Motivated Reasoning," jian, Englewood Cliffs, NJ: Prentice-Hall, 85-123.
Psychological Bulletin, 108 (November), 480-498. Meyers-Levy, Joan and Alice M. Tybout (1989), "Schema
Lazarus, Richard S. (1991), "Progress on a Cognitive-Motiva- Congruity as a Basis for Product Evaluation," Journal of
tional-Relational Theory of Emotion," American Psychol- Consumer Research, 16 (June), 39-54.
ogist, 46 (August), 819-834. Mishra, Sanjay, U. N. Umesh, and Donald E. Stem, Jr. (1993),
Legrenzi, Paolo, Vittorio Girotto, and Philip N. Johnson-Laird "Antecedents of the Attraction Effect: An Information-
(1993), "Focusing in Reasoning and Decision Making," Processing Approach," Journal of Marketing Research,
Cognition, 49 (October-November), 37-66. 30 (August), 331-349.
Levin, Irwin P. and Gary J. Gaeth (1988), "How Consumers Montgomery, Henry (1983), "Decision Rules and the Search
Are Affected by the Framing of Attribute Information Be- for a Dominance Structure:Towards a Process Model of
fore and After Consuming the Product," Journal of Con- Decision Making," in Analyzing and Aiding Decision Pro-
sumer Research, 15 (December), 374-378.
cesses, ed. Patrick C. Humphreys et al., North Holland:
Lucas, RobertE. Jr. (1986), "Adaptive Behavior and Economic
Amsterdam, 343-369.
Theory," Journal of Business, 59 (October, Part2), S401-
Mowen, John C. and James W. Gentry (1980), "Investigation
S426.
of the Preference-Reversal Phenomenon in a New Product
Luce, Mary Frances (1998), "Choosing to Avoid: Coping with
Introduction Task," Journal of Applied Psychology, 65
Negatively Emotion-Laden Consumer Decisions," Jour-
nal of Consumer Research, 24 (March), 409-433. (December), 715-722.
Nedungadi, Prakash (1990), "Recall and Consumer Consider-
, James R. Bettman, and John W. Payne (1997),
"Choice Processing in Emotionally Difficult Decisions," ation Sets: Influencing Choice without Altering Brand
Journal of Experimental Psychology: Learning, Memory, Evaluations," Journal of ConsumerResearch, 17 (Decem-
and Cognition, 23 (March), 384-405. ber), 263-276.
, John W. Payne, and James R. Bettman (1998), Nicosia, Francesco M. (1966), Consumer Decision Processes:
"Trade-Off Difficulty and Choice," working paper, Whar- MarketingandAdvertising Implications, Englewood Cliffs,
ton School, University of Pennsylvania, Philadelphia, PA NJ: Prentice-Hall.
19104. Nowlis, Stephen M. and Itamar Simonson (1997), "Attribute-
Lynch, John G., Jr., Dipankar Chakravarti,and Anusree Mitra Task Compatibility as a Determinant of Consumer Prefer-
(1991), "Contrast Effects in Consumer Judgments: ence Reversals," Journal of Marketing Research, 34
Changes in Mental Representations or in the Anchoring (May), 205-218.
of Rating Scales?" Journal of Consumer Research, 18 Olshavsky, RichardW. (1979), "Task Complexity and Contin-
(December), 284-297. gent Processing in Decision Making: A Replication and
, Howard Marmorstein,and Michael F. Weigold (1988), Extension," Organizational Behavior and Human Perfor-
"Choices from Sets including Remembered Brands: Use mance, 24 (December), 300-316.
of Recalled Attributes and Prior Overall Evaluations," Otnes, Cele, Tina M. Lowrey, and L. J. Shrum (1997), "To-
Journal of Consumer Research, 15 (September), 169- ward an Understandingof Consumer Ambivalence," Jour-
184. nal of Consumer Research, 24 (June), 80-93.
and Thomas K. Srull (1982), "Memory and Attentional Payne, John W. (1976), "Task Complexity and Contingent
Factors in Consumer Choice: Concepts and Research Processing in Decision Making: An Information Search
Methods," Journal of Consumer Research, 9 (June), 18- and Protocol Analysis," Organizational Behavior and Hu-
37. man Performance, 16 (August), 366-387.
(1982), "Contingent Decision Behavior," Psychologi- tude-Behavior Consistency in Children's Responses to

cal Bulletin, 92 (September), 382-402. Television Advertising," Journal of Marketing Research,
, James R. Bettman, and Eric J. Johnson (1988), "Adap- 20 (November), 337-349.
tive Strategy Selection in Decision Making," Journal of Ross, William T., Jr. and Elizabeth H. Creyer (1992), "Making
Experimental Psychology: Learning, Memory, and Cogni- Inferences about Missing Information: The Effects of Ex-
tion, 14 (July), 534-552. isting Information," Journal of Consumer Research, 19
, James R. Bettman, and Eric J. Johnson (1992), "Be- (June), 14-25.
havioral Decision Research: A Constructive Processing Russo, J. Edward (1974), "More Information Is Better: A Re-
Perspective," Annual Review of Psychology, 43, 87-131. evaluation of Jacoby, Speller, and Kohn," Journal of Con-
, James R. Bettman, and Eric J. Johnson (1993), The sumer Research, 1 (December), 68-72.
Adaptive Decision Maker, Cambridge: Cambridge Univer- (1977), "The Value of Unit Price Information," Jour-
sity Press. nal of Marketing Research, 14 (May), 193-201.
, James R. Bettman, and Mary Frances Luce (1996), and BarbaraA. Dosher (1983), "Strategies for Multiat-
"When Time Is Money: Decision Behavior under Opportu- tribute Binary Choice," Journal of Experimental Psychol-
nity-Cost Time Pressure," Organizational Behavior and ogy: Learning, Memory, and Cognition, 9 (October), 676-
Human Decision Processes, 66 (May), 131-152. 696.
, James R. Bettman, and David A. Schkade (1998), and France Leclerc (1994), "An Eye-Fixation Analysis
"Measuring Preferences in a ConstructedWorld: Towards of Choice Processes for Consumer Nondurables," Journal
a Building Code," working paper, Fuqua School of Busi- of Consumer Research, 21 (September), 274-290.
ness, Duke University, Durham, NC 27708-0120. , Victoria Husted Medvec, and Margaret G. Meloy
Peracchio, LauraA. and Alice M. Tybout (1996), "The Moder- (1996), "The Distortion of Information during Deci-
ating Role of Prior Knowledge in Schema-Based Product sions," Organizational Behavior and Human Decision
Evaluation," Journal of Consumer Research, 23 (Decem- Processes, 66 (April), 102-110.
ber), 177-192. , Richard Staelin, Catherine A. Nolan, Gary J. Russell,
Pieters, Rik, Luk Warlop, and Michel Hartog (1997), "The and BarbaraL. Metcalf (1986), "Nutrition Information in
Effect of Time Pressure and Task Motivation on Visual the Supermarket," Journal of Consumer Research, 13
Attention to Brands," in Advances in Consumer Research, (June), 48-70.
Vol. 24, ed. Merrie Brucks and Deborah J. Maclnnis, Schkade, David A. and John W. Payne (1994), "How People
Provo, UT: Association for Consumer Research, 281-287. Respond to Contingent Valuation Questions: A Verbal Pro-
Plott, Charles R. (1996), "Rational Individual Behavior in tocol Analysis of Willingness-to-Pay for an Environmental
Markets and Social Choice Processes: The Discovered Regulation," Journal of Environmental Economics and
Preference Hypothesis," in The Rational Foundations of Management, 26 (January), 88-109.
Economic Behavior, ed. Kenneth J. Arrow et al., New Sethuraman,Raj, CatherineCole, and Dipak Jain (1994), "An-
York: St. Martin's, 225-250. alyzing the Effect of InformationFormat and Task on Cut-
Prelec, Drazen, Birger Wemerfelt, and Florian Zettelmeyer off Search Strategies," Journal of Consumer Psychology,
3 (2), 103-136.
(1997), "The Role of Inference in Context Effects: Infer-
Shafir, Eldar B., Daniel N. Osherson, and Edward E. Smith
ring What You Want from What Is Available," Journal
(1989), "An Advantage Model of Choice," Journal of
of Consumer Research, 24 (June), 118-125.
Behavioral Decision Making, 2 (January-March), 1-23.
Puto, Christopher P. (1987), "The Framing of Buying Deci-
, Daniel N. Osherson, and Edward E. Smith (1993),
sions," Journal of Consumer Research, 14 (December),
"The Advantage Model: A ComparativeTheory of Evalua-
301-315. tion and Choice under Risk," Organizational Behavior and
Ratneshwar, Srinivasan, Comelia Pechmann, and Allan D. Human Decision Processes, 55 (August), 325-378.
Shocker (1996), "Goal-Derived Categories and the Ante-
, Itamar Simonson, and Amos Tversky (1993), "Rea-
cedents of Across-Category Consideration," Journal of son-Based Choice," Cognition, 49 (October-November),
Consumer Research, 23 (December), 240-250. 11-36.
, Allan D. Shocker, and David W. Stewart (1987), "To- Shiv, Baba, Julie A. Edell, and John W. Payne (1997), "Factors
ward Understanding the Attraction Effect: The Implica- Affecting the Impact of Negatively and Positively Framed
tions of Product Stimulus Meaningfulness and Familiar- Ad Messages," Journal of Consumer Research, 24 (De-
ity," Journal of Consumer Research, 13 (March), 520- cember), 285-294.
533. Shugan, Steven M. (1980), "The Cost of Thinking," Journal
Read, Daniel and George Loewenstein (1995), "Diversification of Consumer Research, 7 (September), 99- 111.
Bias: Explaining the Discrepancy in Variety between Com- Simmons, Carolyn J., BarbaraA. Bickart, and John G. Lynch,
bined and Separated Choices," Journal of Experimental Jr. (1993), "Capturing and Creating Public Opinion in
Psychology: Applied, 1 (March), 34-49. Survey Research," Journal of Consumer Research, 20
Richins, Marsha L. (1997), "Measuring Emotions in the Con- (September), 316-329.
sumption Experience," Journal of ConsumerResearch, 24 and John G. Lynch, Jr. (1991), "Inference Effects
(September), 127-146. without Inference Making? Effects of Missing Information
Roedder, Deborah L. (1981), "Age Differences in Children's on Discounting and Use of Presented Information," Jour-
Responses to Television Advertising: An Information-Pro- nal of Consumer Research, 17 (March), 477-491.
cessing Approach," Journal of Consumer Research, 8 Simon, Herbert A. (1955), "A Behavioral Model of Rational
(September), 144-153. Choice," QuarterlyJournal of Economics, 69 (February),
, Brian Sternthal, and Bobby J. Calder (1983), "Atti- 99-118.
(1990), "Invariants of Human Behavior," Annual Re- sions and the Psychology of Choice," Science, 211 (Janu-
view of Psychology, 41, 1-19. ary 30), 453-458.
Simonson, Itamar (1989), "Choice Based on Reasons: The and Daniel Kahneman (1988), "Rational Choice and
Case of Attraction and Compromise Effects," Journal of the Framing of Decisions," in Decision Making: Descrip-
Consumer Research, 16 (September), 158-174. tive, Normative, and Prescriptive Interactions, ed. David
(1990), "The Effect of Purchase Quantity and Timing E. Bell et al., Cambridge: Cambridge University Press,
on Variety-Seeking Behavior," Journal of Marketing Re- 167-192.
search, 27 (May), 150-162. and Daniel Kahneman(1991), "Loss Aversion in Risk-
(1992), "The Influence of Anticipating Regret and less Choice: A Reference-Dependent Model," Quarterly
Responsibility on Purchase Decisions," Journal of Con- Journal of Economics, 106 (November), 1039-1062.
sumer Research, 19 (June), 105-118. , Shmuel Sattath, and Paul Slovic (1988), "Contingent
and Amos Tversky (1992), "Choice in Context: Trade- Weighting in Judgment and Choice," Psychological Re-
off Contrastand Extremeness Aversion," Journal of Mar- view, 95 (July), 371-384.
keting Research, 29 (August), 281-295. and Eldar Shafir (1992), "Choice under Conflict: The
Slovic, Paul (1972), "From Shakespeare to Simon: Specula- Dynamics of Deferred Decisions," Psychological Science,
tions-and Some Evidence-about Man's Ability to Pro- 6 (November), 358-361.
cess Information," Oregon Research Institute Bulletin, 12 and Itamar Simonson (1993), "Context-Dependent
(12). Preferences," Management Science, 39 (October), 1179-
(1995), "The Construction of Preference," American 1189.
Psychologist, 50 (May), 364-371. , Paul Slovic, and Daniel Kahneman (1990), "The De-
, Dale Griffin, and Amos Tversky (1990), "Compatibil- terminants of Preference Reversal," American Economic
ity Effects in Judgment and Choice," in Insights in Deci- Review, 80 (March), 204-217.
sion Making: A Tribute to Hillel J. Einhorn, ed. Robin M. Weber, Elke U., William M. Goldstein, and Sema Barlas
Hogarth, Chicago: University of Chicago Press, 5-27. (1995), "And Let Us Not Forget Memory: The Role of
Stanovich, Keith E. and Richard F. West (1998), "Individual Memory Processes and Techniques in the Study of Judg-
Differences in Rational Thought," Journal of Experimen- ment and Choice," in Decision Making from a Cognitive
tal Psychology: General, 127 (June), 161-188. Perspective, ed. Jerome Busemeyer et al., San Diego, CA:
Academic Press, 33-81.
Sujan, Mita (1985), "Consumer Knowledge: Effects on Evalu-
ation Strategies Mediating ConsumerJudgments," Journal Wedell, Douglas H. (1991), "Distinguishing among Models
of Contextually Induced Preference Reversals," Journal
of Consumer Research, 12 (June), 16- 3 1.
of Experimental Psychology: Learning, Memory, and Cog-
Summers, John 0. (1974), "Less Information Is Better?"
nition, 17 (July), 767-778.
Journal of Marketing Research, 11 (November), 467-
Wernerfelt, Berger (1995), "A Rational Reconstruction of the
468.
Compromise Effect," Journal of Consumer Research, 21
Terry, Deborah J. (1994), "Determinants of Coping: The Role (March), 627-633.
of Stable and Situational Factors," Journal of Personality Wertenbroch, Klaus (1998), "Segmentation and Pricing under
and Social Psychology, 66 (May), 895-910. Consumption Self-Control via Purchase Quantity Ra-
Tetlock, Philip E. (1992), "The Impact of Accountability on tioning," Marketing Science, 17 (4), in press.
Judgment and Choice: Toward a Social Contingency West, PatriciaM., ChristinaL. Brown,andStephenJ. Hoch (1996),
Model," in Advances in Experimental Social Psychology, "ConsumptionVocabularyand PreferenceFormation,"Jour-
Vol. 25, ed. Mark P. Zanna, San Diego, CA: Academic nal of ConsumerResearch, 23 (September), 120-135.
Press, 331-376. Widing, Robert E., III and W. Wayne Talarzyk (1993), "Elec-
, Randall S. Peterson, and Jennifer S. Lemer (1996), tronic Information Systems for Consumers: An Evaluation
"Revising the Value Pluralism Model: Incorporating So- of Computer-Assisted Formats in Multiple Decision Envi-
cial Content and Context Postulates," in The Psychology ronments," Journal of Marketing Research, 30 (May),
of Values: The Ontario Symposium on Personality and 125-141.
Social Psychology, Vol. 8, ed. Clive Seligman et al., Mah- Wilkie, William L. (1974), " Analysis of Effects of Informa-
wah, NJ: Erlbaum, 25-51. tion Load," Journal of Marketing Research, 11 (Novem-
Tversky, Amos (1969), "Intransitivity of Preferences," Psy- ber), 462-466.
chological Review, 76 (January), 31-48. Wright, Peter L. (1974), "The Harassed Decision Maker:Time
(1972), "Elimination by Aspects: A Theory of Pressures, Distractions, and the Use of Evidence," Journal
Choice," Psychological Review, 79 (July), 281-299. of Applied Psychology, 59 (October), 555-561.
(1977), "Features of Similarity," Psychological Re- (1975), "Consumer Choice Strategies: Simplifying vs.
view, 84 (July), 327-352. Optimizing," Journal of Marketing Research, 11 (Febru-
(1996), "Contrasting Rational and Psychological Prin- ary), 60-67.
ciples of Choice," in Wise Choices: Decisions, Games, and Mary Ann Kriewall (1980), "State-of-Mind Ef-
and Negotiations, ed. Richard J. Zeckhauser et al., Boston: fects on the Accuracy with Which Utility Functions Predict
HarvardBusiness School Press, 5-21. Marketplace Choice," Journal of Marketing Research, 17
and Daniel Kahneman (1981), "The Framing of Deci- (August), 277-293.

Constructive Consumer Choice Processes (Bettman Luce Payne)

Uploaded by

Copyright:

Available Formats

Constructive Consumer Choice Processes (Bettman Luce Payne)

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Constructive Consumer Choice Processes (Bettman Luce Payne)

Uploaded by

Copyright:

Available Formats

Constructive Consumer Choice Processes

Author(s): James R. Bettman, Mary Frances Luce, John W. Payne

a majorityof winning (better) attributevalues is retained. TABLE 2

randomchoice. A preference function (reflecting the deci- FIGURE 2

TABLE 3 relational heuristics (such as the componential context

Thus, we hypothesize that relationalreasons will be used correlation,completeness of information,informationfor-

APPENDIX B Comparable versus Noncomparable Choice

Topics for Future Research on Constructive

(1982), "Contingent Decision Behavior," Psychologi- tude-Behavior Consistency in Children's Responses to

You might also like