Procedural Priming and Consumer Judgments:

Effects on the Impact of Positively and

Negatively Valenced Information
HAO SHEN
ROBERT S. WYER JR.*

The cognitive procedure that people use to search for information about a product
is influenced by the ease with which it comes to mind. Unrelated experiences can
activate a search process that governs the order in which favorable and unfavorable
product descriptions are identified and the evaluations that are made on the basis
of them. Five experiments examined the conditions in which these effects occur.
The effects of priming a search strategy on the attention to positively or negatively
valenced information are diametrically opposite to the effects of the semantic (e.g.,
attribute) concepts that are called to mind in the course of activating this strategy.

S uppose consumers wish to evaluate the cost of eating

at a restaurant on the basis of a menu that is posted in
the window. Their judgments can be influenced by different
relatively more expensive in the former case and might be
less inclined to patronize it.
As this example illustrates, consumers' responses to in-
factors. First, the semantic concepts associated with the typ- formation about a product, and their consequent evaluation
ical price can provide a standard to affect their judgment. of it, can be infiuenced by two types of previously acquired
That is, they may evaluate the restaurant to be more ex- knowledge. First, consumers may have declarative knowl-
pensive if they have seen low-priced (vs. high-priced) meals edge about the type of product they are considering (in our
in other places. Moreover, the procedure that they employ example, semantic concepts associated with the typical
to search for information can also infiuence their judgment. price, the type of food being offered, the reputation of the
To form a judgment, they might first identify the most ex- restaurant, etc.). They may use this knowledge to interpret
pensive meal and then, after assessing its attractiveness, con- the information they receive and to make inferences about
sider cheaper ones. Or, they might identify and evaluate the unspecified attributes. Declarative knowledge can also pro-
cheapest meal available before assessing the desirability of vide standards of comparison for evaluating the attractive-
more expensive ones. Often, however, consumers do not ness, quality, or price of the alternatives being considered.
have the time or motivation to consider all of the options In addition, consumers acquire procedural knowledge
available. Then, the judgment they would make if they used that they apply in construing the implications of the in-
the first search procedure is likely to be based on higher formation they receive. These cognitive procedures are em-
prices than the judgment they would make if they used the ployed at several stages of processing. At the judgment
second. Consequently, they might judge the restaurant to be stage, for example, procedural knowledge comes into play
in determining how the implications of different pieces of
information are combined to form an overall evaluation of
*Hao Shen (haoshen@ust.hk) is a doctoral student, and Robert S. Wyer either a single product or a group of products (Anderson
Jr. (mkwyer@ust.hk) is a visiting professor, in the Department of Mar-
keting, Hong Kong University of Science and Technology. The authors
1971; Fishbein 1963; see also Houston, Sherman, and
thank the editor, associate editor, and reviewers, as well as Jaideep Sengupta Baker 1991; Huber, Payne, and Puto 1982; Park and Kim
and Rami Zwick for their comments on an earlier version of the manuscript. 2005). Or, as in our example, it may exert an infiuence at
This research was supported in part by grants HKUST6053/01H, an earlier stage of processing, when consumers seek in-
HKUST6194/04H, and HKUST6192/04H from the Research Grants Coun-
cil, Hong Kong. Correspondence on the manuscript should be sent to Hao
formation to use as a basis for judgment. In either case,
Shen. the judgment can depend on which alternative process hap-
pens to be applied.
John Deighton served as editor and Geeta Menon .served as associate
editor for this artide.
Despite considerable research on the effects oi declarative
knowledge on judgment, the effects of procedural knowl-
Eiectronically pubiished October 5, 2007
edge at the attention stage have not been examined. The
727

© 2008 by JOURNAL OF CONSUMER RESEARCH, Inc. • Vol. 34 • February 2008

All rights reserved. OO93-53O1/2OO8/34O5-OOO8$1O.OO
728 JOURNAL OF CONSUMER RESEARCH

latter effects have theoretical as well as practical implica- by the goal to which the procedure is relevant. However,
tions. For example, negatively valenced information about they can also include extraneous features that happen to
an object has a generally greater impact on judgments than exist in the situation at hand. Therefore, situational factors
positively valenced information (for a review, see Sko- that have nothing to do with the goal can sometimes influ-
wronski and Carlston [1989]). This difference in impact ence which of several goal-relevant procedures is activated
however, has often been attributed to cognitive factors (e.g., and applied. For example, if a consumer wishes to evaluate
differences in the diagnosticity of positive and negative in- the expensiveness of a restaurant, the features of [X] might
formation; see Herr, Kardes, and Kim 1991; Ofir and Si- include a specification of this goal. However, it might also
monson 2001). Motivational factors, such as the desire to include a "consider high values first" feature that happens
avoid negative consequences of a decision (Briley, Morris, fortuitously to be accessible in memory at the time. Thus,
and Simonson 2000, 2005; Briley and Wyer 2002; see also both features in combination may activate a strategy of
Higgins 1997) or to avoid preference-inconsistent infor- searching for high-priced menu entrees before considering
mation (Klein and Ahluwalia 2005), can play a role as well. low ones.
However, the different effects of favorable and unfavorable The productions that constitute procedural knowledge
information could also refiect the use of a more general should be distinguished from descriptions of goal-related
search strategy that leads negatively valenced information activity that are stored in memory as part of declarative
to be identified before positively valenced information and, knowledge. Individuals obviously form mental representa-
therefore, to be used as a basis for judgment when individ- tions of the sequence of steps that are required to attain a
uals are unwilling or unable to integrate the entire set of particular objective (changing a tire, ordering a meal in a
information available. This procedural bias could occur for restaurant, etc.) and may call upon these representations
reasons that are incidental to the product being evaluated. deliberately to facilitate the attainment of this goal. Thus,
The procedure that consumers apply to information at any a man who is unfamiliar with the way to perform a particular
given stage of processing can depend on not only its ap- activity is likely to consult his declarative knowledge for
plicability but also its accessibility in memory (Forster and directions. With practice, however, the efficiency and speed
Liberman 2007; see also Feldman and Lynch 1988; Wyer of performing the sequence of actions involved increase
2008), Its accessibility, in turn, is determined by both the (Schneider and Shiffrin 1977), and the activity can ulti-
frequency and the recency with which it has been used in mately be performed automatically, without consulting de-
the past (Smith 1990; Wyer 2004), Thus, if consumers have clarative knowledge. For example, people who are learning
a chronic disposition to attend to unfavorable features of a to drive a car initially rely on declarative knowledge in
product before considering favorable features, this could determining what to do. With practice, however, they be-
contribute to the negativity bias that has been observed in come able to apply these procedures with a minimum of
previous research. On the other hand, fortuitous experiences cognitive deliberation.
that have led consumers to use a particular procedure in For purposes of this article, we use the term "procedural
responding to the information they receive at one point in knowledge" to refer to the latter procedures (i.e., produc-
time can influence the procedure they apply to the product tions) and not to the sequences of activities that are stored
information they consider in a later, unrelated situation and as part of declarative knowledge. Many studies have not
might decrease or even reverse the negativity bias that would made this distinction. Kirmani, Lee, and Yoon (2004), for
otherwise exist. Five experiments provided evidence that example, showed that priming the concept "if a person puts
this is the case. forth effort into something, it must be worthwhile" influ-
enced participants' evaluation of a product in an unrelated
THEORETICAL BACKGROUND content domain. However, this concept was presumably
stored as part of the declarative knowledge that consumers
Definitions recalled and used as a basis for making inferences about the
products at hand. To this extent, its application did not reflect
Declarative knowledge refers to the content of cognition the use of procedural knowledge as we define it.
(Wyer 2004), In our restaurant example, it could include the
typical price of altemative restaurant offerings and how they
are likely to taste. Procedural knowledge comprises the pro- The Role of Procedural Knowledge in Consumer
cedures that act on declarative knowledge. It includes "the Behavior
sequences of interrelated operations that transform, store,
retrieve, or make inferences based on declarative knowl- Consumer research in several areas provides indirect evi-
edge" (Smith 1994, 101). dence of the effects of procedural knowledge as we con-
Anderson (1983) conceptualized many such procedures ceptualize it. Park and Kim (2005) showed that when choice
as productions, or "if [X], then [Y]" rules, where [X] is a altematives vary along a common set of dimensions, con-
configuration of perceptual or cognitive features and [Y] is sumers often perform a dimension-by-dimension compari-
a sequence of cognitive or motor acts that are elicited au- son of their features and invoke criteria similar to those
tomatically when the precondition [X] is met. The features postulated by Simonson (1989). However, if consumers are
that compose this precondition may be determined in part asked to evaluate each choice altemative separately before
PROCEDURAL PRIMING 729

stating their preference, they base their preference on a com- procedure was diametrically opposite to the effect of the
parison of these overall evaluations. Thus, the procedure declarative knowledge that had been activated by the prim-
that participants used to compute their preferences depends ing task. Experiment 4 confirmed these effects and showed
on the ease with which it can be performed. that they combined additively to affect judgments. Finally,
In a quite different domain, Dhar, Huber, and Khan (2007) experiment 5 demonstrated that the effect of priming pro-
showed that inducing consumers to make a purchase at the cedural knowledge generalized to the processing of infor-
beginning of an experiment increased their likelihood of mation about objects that were quite different from those
making a second purchase later. Xu and Wyer (2008) to which the priming task pertained.
showed that simply asking consumers to report their pref-
erence for choice altematives in one domain increased their
willingness to make a purchase in a second, unrelated do-
main. The need to state a preference in one domain appar- EXPERIMENT 1
ently activated a comparative judgment process that induced
consumers to think about which product they wanted to buy The first experiment provided preliminary evidence that
in the second domain without considering the option of not the search procedure that participants use in the course of
buying anything at all. Another example was provided by performing a task in one situation can persist to influence
Briley and Wyer (2002), who showed that feelings of group their processing of information in a different situation and
membership induced a disposition to compromise that, once the judgments they make on the basis of it, Shafir (1993)
activated, was applied in totally unrelated product choice found that consumers focus their attention on favorable at-
situations. tributes when they consider whether to choose a product,
but focus on unfavorable attributes when deciding whether
to reject it. If this is so, asking participants to choose or
The Impact of Negatively Valenced Information reject a product in one situation may induce a search strategy
The tendency for negatively valenced information to have that, once activated, will generalize to other situations they
a greater impact on judgments than positively valenced in- encounter subsequently.
formation has been commonly identified in both psychology
(Bimbaum 1974; Skowronski and Carlston 1989) and con-
sumer research. For example, people may expect descrip-
tions of a person or product to be favorable, so unfavorable Method
(unexpected) information may be more attention-getting
(Fiske 1980) or more diagnostic (Skowronski and Carlston Design and Subjects. Eighty-five Hong Kong univer-
1989), Similarly, consumers may expect recommendations sity students were randomly assigned to conditions of a 2
of a product or service to be biased in a socially desirable (priming task: choose vs. reject) x 2 (time pressure: high
way, so they perceive a negative evaluation to have greater vs, low) between-subjects design.
credibility (Basurcy, Chatterjee, and Ravid 2003; Chevalier
and Mayzlin 2006), Procedure. To introduce the study, participants were
Other factors can also play a role. Low risk takers, or told that the experimenter was interested in how people
consumers with a prevention focus (Higgins 1997), may be make purchase decisions on the basis of product descrip-
motivated to avoid negative consequences of a decision re- tions. On this pretense, they were exposed to descriptions
gardless of the benefits it might provide, and might weight of five products (a job offer, a hotel, a room, an MP3, and
negative attributes more heavily for this reason (Briley et an automobile), each consisting of four attributes. Some
al. 2000, 2005; Briley and Wyer 2002), participants were told to decide whether they would choose
The different effects of negatively and positively valenced each of the five products by circling either "yes" or "no,"
information identified in the aforementioned research, how- Others were told to indicate whether they would reject each
ever, were typically the result of processing that occurred product by circling one of these alternatives.
at the time that different pieces of information are integrated After participants had completed this task, the experi-
to form a judgment. As our previous example suggests, menter passed out an ostensibly unrelated shopping ques-
however, cognitive procedures that are performed at the in- tionnaire with instructions that attempted to glean how con-
formation seaich stage can bias the subset of information sumers make judgments about products when they go
that is identified and used as a basis for judgment. The shopping. Participants read descriptions of a personal com-
present research investigated the effects of this bias. Ex- puter consisting of five positive attributes and five negative
periments 1 and 2 provided evidence that priming consumers ones (see table 1). After reading the information about the
to focus their attention on attributes with high or low values computer, they evaluated it along a scale from 0 (not at-
in one situation can influence the procedure they use to tractive at all) to 10 (very attractive). Under high time pres-
search for information in a second, unrelated situation, and sure conditions, participants were told they would have less
that the magnitude of this influence is greater when they than 15 seconds to evaluate the computer. In low time pres-
have little time to think about the information. Experiment sure conditions, they were told they could have as much
3 showed that the effect of priming an information search time as they liked.
730 JOURNAL OF CONSUMER RESEARCH

TABLE 1 high importance. These procedures biased the end of the

DESCRIPTIONS OF TARGET PRODUCTS
rating scale that participants used in reporting judgments in
a later task, producing anchoring-and-adjustment effects of
Experiment 1
the sort identified by Tversky and Kahneman (1974).
Experiments 1 and 2
(personal computer) (travel destination) We expected that the ranking task would also influence
the procedure that participants use at an early stage of pro-
High RAM capacity Beautiful scenery cessing. Rank ordering stimuli from high to low requires
Small keyboard Pollution problem participants to identify high values before considering low
Poor sound quality Plenty of nightspots ones, whereas rank ordering from low to high requires them
Little software included Good theaters
Protective screen Expensive to identify low values before high ones. If participants who
High CPU speed Attractive beaches have rank ordered stimuli using one of these procedures are
Insensitive mouse Possible bad weather later required to consider several pieces of information in
Installment payments Overcrowded order to make a judgment, the ranking procedure they used
Sold in reputable store Good museums
Unattractive keyboard color Long travel time earlier should influence the pieces they are most likely to
identify and use. Furthermore, this influence should be par-
ticularly evident when the time available to process this
Results information is limited.
Participants were expected to identify favorable attrib-
utes before considering unfavorable ones in the course of
Method
deciding whether to choose products in the priming task, One hundred and twenty-three Hong Kong university stu-
but to identify unfavorable attributes before favorable ones dents were randomly assigned to the conditions of a 2 (rank-
in the course of deciding whether to reject them. Further- ing procedure) x 2 (time allowed for evaluating products)
more, these search strategies were expected to generalize design. The type of product evaluated (computer vs. travel
to the product evaluation task they performed subse- location) was varied within subjects.
quently. When they did not have an opportunity to consider To introduce the study, the experimenter told participants
all of the attributes presented, therefore, participants were that the Hong Kong Vocational Testing Service was devel-
expected to evaluate the product more favorably in the first oping an ordering test for use in job placement. Participants
condition than in the second. When they had as much time were exposed to descriptions of five stimuli identical to those
as they wanted, however, this difference was expected to used in experiment 1. Each description consisted of four
be minimal. attributes. Some participants were told to rank each set of
Our expectation was confirmed. Participants under time attributes from most attractive to least attractive, whereas
pressure evaluated the personal computer more favorably if others were told to rank them from least to most attractive.
they had decided whether to choose stimuli (M = 5.80) than Explicit instructions were given concerning the procedure
if they had decided whether to reject stimuli (M = 4,64; to use. For example, subjects in the first condition were told
F(l, 81) = 3.76, p < .06). However, the reverse was true in the following. "First look for the attribute that is most at-
the absence of time pressure (4.67 vs, 5.68, respectively; tractive to you and write the attribute name in the first box.
F(l, 81) = 2.94, p < .09). The interaction of time pressure Then look for the next most attractive attribute and put it
and priming was significant (F(l, 81) = 6,67, p < .01), in the second box. Continue this manner until all attributes
are ranked."
EXPERIMENT 2 When participants had finished the ordering test, they
were asked to perform the ostensibly unrelated "shopping
Experiment 1 confirmed the hypothesis that performing questionnaire" using instmctions identical to those em-
an ostensibly unrelated product choice task can prime an ployed in experiment 1. In this study, however, they eval-
information search procedure and that this procedure, once uated both a personal computer and a vacation spot, each
activated, is applied in subsequent situations as well. Con- based on 10 attributes (see table 1). They were again told
sequently, it biases the order in which favorable and unfa- they would have either 15 seconds or as much time as they
vorable attribute information is considered and influences wanted to evaluate each product. Finally, after making their
the relative impact of these attributes on judgments when evaluations, they were asked to recall two attributes of each
the time to process all of the available information is limited. product they had considered.
Experiment 2 confirmed this conclusion using a different
priming procedure. In addition, we obtained supplementary
recall data to confirm our assumptions about the type of Results
information to which participants attended. Product Evaluation. Attractiveness ratings of the two
Specifically, we used a priming task similar to that em- products are shown in table 2 as a function of ranking
ployed by Schwarz and Wyer (1985). Participants in procedure, time pressure, and product type. The effects of
Schwarz and Wyer's research rank ordered a set of social these variables confirmed the results of experiment 1, The
issues either from high to low importance or from low to interaction of procedure and time pressure was significant
PROCEDURAL PRIMING 731

TABLE 2 formation extensively. This finding confirms our assumption

MEAN REACTIONS TO PERSONAL COMPUTER AND TRAVEL
of the sequential search process that is activated by this
LOCATION AS A FUNCTION OF PROCEDURE AND TIME procedure. It is worth noting, however, that when partici-
PRESSURE—EXPERIMENT 2 pants had ample time to process all of the information pre-
sented, their evaluations were more favorable when partic-
Personal computer Travel location ipants had rank ordered from least to most favorable. That
Time No Time No
is, the last (most recent) information that participants pro-
Product evaluation pressure pressure pressure pressure cessed had the greatest effect. Although this reversal was
unexpected, it is consistent with the effects observed by
Best to worst 3,69 3.94 4.90 4.41 Diehl and Zauberman (2005). That is, they also found pri-
Worst to best 3.27 4.63 3.67 5.22 macy effects of information presentation order when recip-
.42 -.69 1.23 -.81 ients' cognitive resources were limited, but recency effects
when they were not.
{F{ l,119) = 7.55,p<.01) and was not contingent on prod-
uct type {F < 1). Specifically, participants under time pres- EXPERIMENT 3
sure evaluated products more favorably if they had rank
ordered attributes from most to least attractive (M = 4.29) Experiments 1 and 2 demonstrated the effects of priming
than if they had ordered them from least to most attractive procedural knowledge on participants' processing of favor-
{M = 3.47; F(l, 119)= 4.00, p < .05). In the absence of able and unfavorable stimulus information in a later situa-
time pressure, however, the reverse was true (4,17 vs. 4.92, tion; the effect of declarative knowledge on the attention to
respectively; F{1,119) = 3.57, p < .06), In this case, there- this information was not clear. In principle, declarative
fore, the last (most recent) information that participants con- knowledge could have at least two different effects under
sidered had the greatest effect. the conditions we examined. First, the attribute concepts
made salient by the rank-ordering task employed in exper-
Attribute Recall. Our interpretation of the judgment iment 2 could affect participants' interpretations of the prod-
data assumes that participants under time pressure were un- uct information they later consider and use as a basis for
able to process all of the attributes. If this is so, the attributes judgment. These effects are unlikely, however, unless the
they can recall should be the ones they had considered first. information exemplifies the specific attribute concepts that
When participants were given enough time to read all of are primed (Higgins, Rholes, and Jones 1977), It seemed
the attributes presented, however, this tendency should be more likely, therefore, that the primed attribute concepts
less evident. would provide comparative standards relative to which the
To evaluate this possibility, the two attributes that partic- favorableness of the attribute information is evaluated, pro-
ipants recalled of each product were coded as either 1 (if ducing contrast effects on judgments of the sort identified
favorable) or — 1 (if unfavorable), and incon^ect responses by Adaval and Monroe (2002; see also Herr 1986; Ostrom
were scored as 0. These scores were summed over the two and Upshaw 1968), Although these effects were not evident
products and analyzed as a function of time pressure, ranking in experiments 1 and 2, they are more likely to occur when
procedure, and product type. The favorableness of the at- the attribute dimension made salient by the priming task and
tributes recalled by participants under time pressure was the dimension of judgment are the same.
greater if they had rank ordered attributes from most to least Experiments 3 and 4 provided an indication of the role
attractive than if they had rank ordered them from least to of declarative knowledge activation as well as procedural
most attractive (0,66 vs, 0.13, respectively; F(l,119) = knowledge activation. Participants in experiment 3 were
3.93, p < .05). However, this difference was in the opposite given two sets of conditions. In procedural knowledge
direction when participants were not under time pressure priming conditions, they were given the prices of 10 hotels
(0.20 vs. 0,56, respectively; F(l, 119) = 2.02,p> .10), The and asked to rank order them from either highest to lowest
interaction of ranking procedure and time pressure was sig- or lowest to highest. Then, in the judgment task, they were
nificant (F(l, 119) = 5.80, p < ,02) and was not contingent given a list of the prices of hotels in a different city and
on product type {F < 1). asked to evaluate their average cost. The procedural knowl-
edge activated by the ranking task should dispose partic-
ipants to judge the hotels as more expensive when they
Discussion had ranked from high to low than when they had ranked
The first two experiments confirmed our hypothesis that from low to high. However, the prices made salient by the
activating a search procedure in the course of performing ranking task could also activate general price-related con-
one task generalizes to an ostensibly unrelated judgment cepts that are used as standards of comparison in evaluating
task that participants perform subsequently, biasing the in- the cost of the target hotels. To this extent, contrast effects
formation they are most likely to consider and the evalua- might occur of the sort identified by Adaval and Monroe
tions that they base on this information. As we expected, (2002), and these effeets could offset the effects of pro-
the effects of the search strategy were most significant when cedural priming.
participants did not have the time to process all of the in- It was therefore necessary to establish that the effect of
732 JOURNAL OF CONSUMER RESEARCH

declarative knowledge would in fact be evident under con- pants to rank prices, all participants were asked to identify
ditions in which the procedural knowledge that participants the city whose hotels were highest in average quality. This
employed was controlled. Participants in declarative knowl- procedure was assumed to activate a concept of high prices
edge priming conditions were given a list of either the five in one condition but a concept of low prices in the other
highest priced hotels or the five lowest priced hotels that condition.
participants considered in procedural knowledge priming Then, participants were asked to assume that they were
conditions. These stimuli were likely to prime concepts sim- planning to visit Dalian, a city in China, and were searching
ilar to those that participants might activate in the course for a hotel on a Web site. They were told to read over the
of performing the rank-ordering task without activating a prices of the hotel rooms listed and to evaluate the average
search procedure. That is, the stimuli should activate a con- price of hotel rooms in the city. Six hotels were listed in
cept of either high-priced hotels or low-priced hotels that low information load conditions and 24 in high information
participants use as a standard of comparison in evaluating load conditions. In each case, the prices ranged from $15
the cost of the target hotels that they consider subsequently. to $330, but most were moderate (e.g., $60-$80) and there-
As a result, participants should judge the targets to be less fore were generally dissimilar to both the high prices and
expensive when the price of hotels in the priming task was the low prices to which participants were exposed in the
high than when it was low. priming task.
To provide a further test of the different effects of pro- After receiving this information, participants turned over
cedural and declarative knowledge priming, we again ma- the page and estimated the average price of the hotel rooms
nipulated participants' abilities to process all of the infor- in the city along a scale from 0 (not expensive at all) to 10
mation available for use as a basis of their impression (very expensive). (They were instructed not to refer back
judgments. Rather than varying the time that participants to the previous page when making their ratings.) They also
were given to make their judgments, however, we manip- indicated the price they were willing to pay for a hotel room
ulated the number of hotels in the list that participants were in that city. To check on the procedure that participants used
given to consider. The effect of procedural priming should in searching prices, they were also asked whether they first
be evident primarily when the amount of information that paid attention to the highest prices or to the lowest prices
participants were required to search was high. However, the when they made the judgment.
use of primed concepts as a standard of comparison occurs
at a later stage of processing. Thus, if anything, these effects
should be more evident when processing load is low and Results
participants have more opportunity to engage in these com- Manipulation Check. Participants were more likely to
parison processes. report paying attention to high prices before considering low
prices when they had been primed to rank order prices from
Method highest to lowest (83%) than when they were primed to rank
order prices from lowest to highest (56%; Wald x^ =
One hundred and twenty-four Hong Kong university stu-
4.40, p < .03), In declarative knowledge priming conditions,
dents took part in this experiment, half of whom participated
however, this difference was not apparent (80% vs, 78%),
under procedural knowledge priming conditions and the rest
of whom participated under declarative knowledge priming Price Evaluation. Estimates of the expensiveness of the
conditions. Participants in the former conditions were ran- target hotels are summarized in the top section of table 3,
domly assigned to four combinations of ranking procedure The effects of declarative knowledge priming in the absence
(high to low vs, low to high) and the amount of information
to be processed (large vs. small). Participants in the latter
conditions were assigned to four combinations of declarative TABLE 3
knowledge activation (high price vs. low price) and amount
MEAN REACTIONS TO PRICES OF HOTEL ROOMS
of information. AS A FUNCTION OF KNOWLEDGE, TYPE OF KNOWLEDGE,
To introduce the study, the experimenter told participants AND PROCESSING LOAD—EXPERIMENT 3
that the Hong Kong Vocational Testing Service was devel-
oping a battery of tests for use in job placement. In pro- Procedural knowledge Declarative knowledge
cedural knowledge priming conditions, participants were ex- priming pnming
posed to prices of rooms available in 10 hotels in each of High load Low load High load Low load
three cities. The prices ranged from $33 to $256. Participants
were asked to rank the prices in each city either from highest Price evaluation
to lowest or from lowest to highest. In declarative knowledge High to low 5,63 4,18 4,18 3.87
Low to high 4,15 4,33 5.06 4.71
priming conditions, participants only saw five prices of hotel 1,48 -,15 -.88 -.84
rooms in each city. In some conditions, however, these were Willingness to pay:
the five highest prices in the list that was presented under High to low $87 $75 $71 $78
procedural priming conditions, and in other cases, they were Low to high $68 $98 $89 $80
the five lowest prices in the list. Instead of asking partici- Wdiff $19 -$23 -$18 -$2
PROCEDURAL PRIMING 733

of procedural knowledge priming were as expected. That EXPERIMENT 4

is, participants in declarative knowledge priming conditions
evaluated the average price as less expensive if they were Although an effect of declarative knowledge priming was
exposed to high prices {M = 4.03) than if they were ex- not evident under the procedural knowledge priming con-
posed to low prices (M = 4.90; F(l, 116) = 3,88,/? < ,05), ditions of experiment 3, this does not indicate that the de-
Furthermore, this was true independent of the amount of clarative knowledge activated by the priming task had no
information to which they were exposed. effect at all. Although the ranking procedure may have in-
However, the effects of declarative knowledge were not fluenced the order in which the prices were considered, par-
evident in procedural knowledge priming conditions. When ticipants may ultimately have attended to all of the prices
a large amount of price information was provided, partici- they ranked. Consequently, a contrast effect of these prices
pants in these conditions evaluated the average price as more on judgments of the sort that results from declarative knowl-
expensive if they had rank ordered prices from highest to edge activation might have occurred that was similar in both
lowest {M = 5.63) than if they had ranked them from lowest ranking conditions.
to highest (yW = 4.15; F(l, 116) = 7.15, p < ,01), This in- Experiment 4 provided a more direct test of the combined
fluence disappeared, however, when the amount of price effects of procedural and declarative knowledge priming.
information was small (4,33 vs. 4.18, F < 1). The interaction Participants in moderate price priming conditions rank or-
was marginally significant {F{1,116) = 3.22, p < .08) and dered the price of the hotels employed in priming conditions
is consistent with our hypothesis. of experiment 3. In high price priming conditions, however,
the price of each hotel in the priming task was increased by
Willingness to Pay. The price that participants were HK$300, The effect of the rank-ordering task on the ac-
willing to pay for a hotel was affected in much the same cessibility of procedural knowledge was expected to be iden-
manner as their expensiveness judgments, as shown in the tical to that observed in experiment 3. However, if the effects
second section of table 3. In declarative knowledge priming of declarative knowledge priming occur also, the magnitude
conditions, exposure to high prices slightly decreased par- of the prices to which participants were exposed in the prim-
ticipants' willingness to pay (mean difference = -$10, ing task should induce a contrast effect in both ranking
p> .10), and this was true regardless of the amount of price conditions. That is, participants should judge the target ho-
information they considered. However, priming a disposition tels to be less expensive when high-priced hotels were
to search for high prices before low prices increased the ranked than when moderately priced hotels were ranked,
price that participants were willing to pay for the hotel (mean A second objective of this study was to establish the
difference = $19; F{\, 116) = 2.76, p < ,10) when a large robustness of procedural priming effects. In the first three
amount of information was presented. However, it nonsig- experiments, the stimuli presented in the judgment task were
nificantly decreased their willingness to pay when the num- ordered randomly. It might be reasonable to suppose that
ber of prices was small (mean difference = - $ 2 3 ; p > the effect of procedural priming would be particularly evi-
,10), The interaction of priming procedure and amount of dent in such conditions, Diehl and Zauberman (2005) found
information was significant (F(l, 116) = 5.09, p < .03). that when participants' cognitive resources were limited,
they estimated the average price of hotels to be higher if
the prices were presented in descending order than if they
Discussion were presented in ascending order. One might speculate that
The results of this experiment demonstrated the different if the stimuli were already presented in order of magnitude,
effects of procedural knowledge priming and declarative the search strategy that people employ would be determined
knowledge priming. That is, participants who were exposed by this order and that priming procedural knowledge in an
to high or low prices without having performed the rank- unrelated situation would have little effect. In fact, however,
ordering task appeared to use a concept activated by these presentation order had no effect at all.
prices as a standard of comparison when evaluating the av-
erage price of hotel rooms in the subsequent task, and there-
fore evaluated it as less expensive when the prices they
Method
encountered in the first task were high compared to when Ninety-three Hong Kong university students participated.
they were low. Participants were first asked to rank the prices of 10 hotels
Priming procedural knowledge had quite different effects. in each of three cities either from highest to lowest or from
As in our earlier studies, participants who were primed to lowest to highest. In moderate price priming conditions, the
use a particular search procedure in the rank-ordering task prices they ranked were the same as those in experiment 3.
appeared to follow the same procedure in identifying in- In high price priming conditions, however, each price that
formation in the subsequent judgment task. Thus, when they participants encountered was increased by HK$300. (For
were given a large amount of price information to consider, example, prices of $62 and $75 in moderate price priming
they evaluated the average price of the hotels as more ex- conditions were replaced by prices of $362 and $375, re-
pensive if the procedural priming task required attention to spectively.) Participants then estimated the average price of
high prices. When they considered only a few prices, how- the 24 hotels they considered in the high information load
ever, this effect disappeared. conditions of the third experiment. However, the prices were
734 JOURNAL OF CONSUMER RESEARCH

not ordered randomly but, rather, were presented in either TABLE 4

descending order (from highest to lowest) or ascending order
ESTIMATES OF EXPENSIVENESS AS A FUNCTION OF
(from lowest to highest). After receiving the information, RANKING PROCEDURE, PRiMING PRICE MAGNITUDE, AND
participants answered the same questions asked in experi- INFORMATION PRESENTATION ORDER—EXPERIMENT 4
ment 3 and, as an additional manipulation check, evaluated
the prices they had ranked in the priming task along a scale High prices Low prices
from 0 (not at all expensive) to 10 (very expensive). presented first presented first
High Moderate High Moderate
Estimates of price price price price
Results expensiveness primes primes primes primes
Manipulation Checks. Our manipulations were suc-
High to low 4,62 5.91 5.18 6.27
cessful. Participants rated the prices they had ranked in the Low to high 4,00 5.25 4.45 4.92
first task as more expensive if the prices were high (M = ,62 .66 .73 1.35
6.74) than if they were low {M = 4.35; F(l,85) = 33,84,
p < .0001). Furthermore, they were more likely to report
paying attention to high target prices before low ones when judgment-relevant product information appear to exert an
they had previously ranked prices from high to low than influence independently of the order in which this infor-
when they had ranked them from low to high (70% vs. 43%; mation is presented. In fact, they may override any effects
Wald x^ — 6.69, p < .01), Their attention to high prices was that presentation order might otherwise have. Participants
also greater when the prices were presented in descending reported being more likely to consider the first information
order rather than ascending order (83% vs. 28%; Wald presented before other information. Despite this fact, they
X^ = 24,54, p< ,0001). Thus, participants' search of the apparently searched the list for the type of information that
information in the judgment task was influenced by the order was relevant to the procedure that had been primed earlier,
in which the information was presented as well as the prim- and took this information into account before assessing the
ing procedure. implications of other information.

Price Estimates. Participants' estimates of the cost of EXPERIMENT 5

hotels they considered in the judgment task are summarized
in table 4. Participants judged the average price of hotels Research in other stimulus domains suggests that the ef-
to be more expensive if they had rank ordered prices in the fect of activating a procedure in the course of processing
priming task from high to low (M = 5.46) than if they had one type of information can generalize to quite different
ranked them from low to high (M = 4.68; F(l,85) = types of stimuli and judgment situations (Schwarz and Wyer
4.26, p < .04). This confirms the conclusions drawn from 1985; Smith, Branscombe, and Bormann 1988), To evaluate
experiment 3, However, an effect of declarative knowledge the applicability of our findings to consumer judgment sit-
priming was also evident. That is, participants evaluated the uations outside the laboratory, however, it seemed desirable
average price as less expensive if the prices to which they to establish the generalizability in the product domain as
were exposed in the priming task were high (M = 4,57) well. Experiment 5 accomplished this. In addition, it pro-
than if they were moderate {M = 5,55; F(l,85) = 6.39, vided evidence concerning the persistence of the effects of
p < .01), The interaction of the ranking procedure and prim- procedural knowledge priming over time.
ing price magnitude was not significant {p > ,10), indicating
that the effects of procedural knowledge priming and de-
clarative knowledge priming were independent. Further- Method
more, the order in which the prices were presented in the The experiment was a 2 (ranking procedure) x 2 (pro-
judgment task had no effects whatsoever (F< 1). cessing load) X 2 (evaluation delay) between-subjects de-
sign. Ninety-seven Hong Kong university students partici-
Discussion pated. Participants were told that they would perform a test
of their ability to rank order numbers quickly and accurately.
Experiment 4 eliminated two ambiguities conceming the Participants were told that three groups of students had taken
generalizability of the effects of priming procedural knowl- an exam and that a list was compiled of the number of
edge. First, these effects do not override the impact of de- correct answers made by students in each group. They were
clarative knowledge accessibility on judgments. Rather, the instructed to rank the students in each group in terms of the
activities in which consumers engage before they make a number of correct answers they attained. Some participants
product judgment may simultaneously activate both pro- ordered the scores from highest to lowest and others ordered
cedural knowledge and declarative knowledge, and the ef- them from lowest to highest.
fects of the two types of knowledge are independent and Then, participants received a task that was described as
additive. a measure of their ability to interpret numerical data. They
On the other hand, the information search procedures that were told to assume {a) that they were a marketing manager
people have accessible in memory before they encounter whose company planned to launch a new advertisement, {b)
PROCEDURAL PRIMING 735

that in order to see whether consumers like this advertise- TABLE 5

ment or not, they have obtained ratings of the ad by some
MEAN IMPRESSION OF CONSUMER RATINGS AS A
consumers, and (c) that they were to read over the ratings FUNCTION OF PROCEDURE, PROCESSING LOAD,
and form an overall impression of the ad on the basis of AND EVALUATION DELAY—EXPERIMENT 5
them. Participants were then exposed to consumer ratings
of an ad that ranged from 1 to 7. In low processing load Immediate
conditions, participants reviewed 16 ratings, four of which judgment Delay
were in the middle scale category (4); the others were dis- Impression of
tributed evenly over the remaining categories. In high pro- consumer High Low High Low
cessing load conditions, they reviewed 160 ratings, 40 of ratings load load load load
which were in the middle category; the others were evenly
distributed over the other categories. Thus, the mean rating High to low 1,60 ,54 2,00 ,25
Low to high ,27 ,31 ,85 1,15
was 4 in each case.
1,33 ,23 1,15 -,90
After reviewing this information, participants in imme-
diate-judgment conditions tumed over the page and evalu-
ated the ad along a scale from - 5 (very bad) to +5 (very
good). In contrast, participants in delay conditions were sumers spontaneously employ when searching for infor-
given another unrelated task and did not evaluate the ad mation at the time it is presented. Moreover, the negativity
until 10 minutes later. bias that results from processing at the computation stage
can sometimes be eliminated or even reversed by fortuitous
experiences that influence the order in which the information
Results is identified at the time it is first received.
The effect of procedural priming on ad evaluations was Experiment 1, for example, provided evidence that simply
analogous to that observed in previous experiments. As table asking participants to decide whether they would choose to
5 shows, participants in high information load conditions buy each of a set of products disposed them to search for
evaluated the ad more favorably if they had ranked numbers favorable attributes before unfavorable ones in an unrelated
from highest to lowest {M = 1.82) than if they had ranked product evaluation situation. Similar effects were evident in
them from lowest to highest {M = 0.58; F(l,89) = 7.64, experiment 2 as a result of rank ordering product attributes
p < .01). In low information load conditions, however, this from most to least favorable. Once activated, these dispo-
difference was not evident {F < 1). The interaction of rank- sitions influenced the order in which attributes of a target
ing procedure and processing load was significant product were identified in an unrelated situation. Thus, they
(F(l,89) = 6.52,/7 < ,01). However, neither this interaction influenced evaluations of the product when participants had
nor any other effects were significantly contingent on delay little time to process all of the information presented. Ex-
conditions {p>.lO). Therefore, the effects of procedural periments 3 and 4 showed similar effects on the overall
knowledge activation on judgments appear to persist over evaluation of a group of objects. They further demonstrated
time, that the effects were not restricted to judgments of favor-
ableness alone, but generalized to more specific attributes
that had evaluative implications (price). Finally, experiment
GENERAL DISCUSSION 5 demonstrated that the effects of activating procedural
Consumers are often called upon to evaluate a product knowledge in the course of performing a task in one domain
based on several pieces of information about both the same are not restricted to processing of stimuli in this domain but
product and others. The manner in which they process this generalize to other, quite different types of stimuli as well.
information, and therefore, the judgments they make on the The search strategies activated by procedural knowledge
basis of it, may be influenced by objectively unrelated cog- priming were applied independently of the order in which
nitive activities in which they engage before encountering the information was presented. Although the information in
the information. Furthermore, these activities can influence experiment 4 was presented in either an ascending or a
not only the product-relevant knowledge that the consumers descending order, this difference did not affect the search
bring to bear on their constmal of the information's impli- strategy that participants employed in construing its impli-
cations but also the procedure they use to search and identify cations, (If anything, it might have increased the ease with
features of this information. The effect of this procedure is which the primed search procedure could be applied,) These
reflected in the judgments they make. effects, however, were only pronounced when participants
The studies reported in this article have more specific were unable to process all of the information they received.
implications for the factors that influence the relative impact In fact, when participants had time to consider all of the
of positively and negatively valenced product information. information they had available, the most recently processed
That is, the relatively greater influence of negatively va- information appeared to have the greatest impact (see ex-
lenced information can result not only from the greater periments 1 and 2).
weight that this information is given in the course of com- The rank-ordering procedure we employed in experiments
puting a judgment but also from the procedure that con- 2-5 was likely to activate declarative knowledge about the
736 JOURNAL OF CONSUMER RESEARCH

stimuli being ranked. However, the effects of activating de- Anderson, Norman H, (1971), "Integration Theory and Attitude
clarative and procedural knowledge were independent and Change," Psychological Review, 78 (3), 171-206,
additive. Experiments 3 and 4 demonstrated that the de- Basuroy, Suman, Subimal Chatterjee, and Abraham S, Ravid
clarative knowledge that was likely to be activated by the (2003), "How Critical Are Critical Reviews? The Box Office
Effects of Film Critics, Star Power, and Budgets," Journal of
rank-ordering task provided a standard of comparison that
Marketing, 67 (October), 103-17,
participants used to evaluate the implications of the infor-
Bimbaum, Michael H, (1974), "The Nonadditivity of Personality
mation they received in the judgment task, producing con- Impressions," Journal of Experimental Psychology (Mono-
trast effects. graph), 102 (3), 543-61,
The priming procedure we used in these experiments was Briley, Donnel A,, Michael W, Morris, and Itamar Simonson
somewhat artificial. As experiment 1 showed, however, ex- (2000), "Reasons as Carrier of Culture: Dynamic versus Dis-
periences that consumers are more likely to encounter out- positional Models of Cultural Influence on Decision Making,"
side the laboratory can have similar effects. For example, Joumal of Consumer Research, 27 (September), 157-78,
a desire to maximize the positive consequences of a decision (2005), "Cultural Chameleons: Biculturals, Conformity
might stimulate a disposition to search for stimuli with fa- Motives, and Decision Making," Journal of Consumer Psy-
vorable attributes, which are of particular relevance to this chology, 15 (4), 351-62,
objective, whereas a desire to minimize the negative con- Briley, Donnel A, and Robert S, Wyer (2002), "The Effect of Group
sequences of a decision might stimulate a disposition to Membership Salience on the Avoidance of Negative Out-
search for unfavorable attributes first. These dispositions comes: Implications for Social and Consumer Decisions,"
could activate search procedures similar to those that re- Journal of Consumer Research, 29 (December), 400^15,
sulted from performing the rank-ordering task. Search strat- Chevalier, Judith A, and Dina Mayzlin (2006), "The Effect of Word
of Mouth on Sales: Online Book Reviews," Joumal of Mar-
egies can also depend on the type of products being judged.
keting Research, 43 (August), 345-54,
For example, consumers might always search cheap prices Dhar, Ravi, Joel Huber, and Uzma Khan (2007), "The Shopping
first when they see yogurt or fmit, or they might search Momentum Effect," Joumal of Marketing Research, 44 (Au-
negative information first if they have a quality control goal. gust), 370-78,
Once these strategies are activated, however, they might be Diehl, Kristin and Zauherman Gal (2005), "Searching Ordered
applied automatically in other situations in which these goals Sets: Evaluations from Sequences under Search," Joumal of
are no longer relevant. Consumer Research, 31 (March), 824-32,
Some qualifications on our conclusions are worth noting. Feldman, Jack M, and John G, Lynch Jr, (1988), "Self-Generated
For example, participants in the present experiments were Validity and Other Effects of Measurement on Belief, Atti-
given limited opportunity to practice the rank-ordering test. tude, Intention, and Behavior," Joumal of Applied Psychol-
Past research (Smith 1990) suggests that procedural knowl- ogy, 73 (3), 421-35,
edge is slow to develop but that once it is acquired, its effects Fishbein, Martin (1963), "An Investigation of the Relationships
persist for some time. Even in the conditions we constructed, between Beliefs about an Object and the Attitude toward That
Object," Human Relations, 16 (August), 233-40,
the effects of priming procedures were evident after a short
Fiske, Susan T. (1980), "Attention and Weight in Person Percep-
delay. Perhaps with more practice, the effects would persist tion: The Impact of Negative and Extreme Behavior," Joumal
for a much longer period of time, Kolers (1976), for ex- of Personality and Social Psychology, 38 (6), 889-906.
ample, found that practice reading text in novel orientations Forster, Jens and Nira Liberman (2007), "Knowledge Activation,"
(e,g., right to left, backwards) improved participants' per- in Social Psychology: Handbook of Basic Principles, 2nd ed.,
formance on this task 1 year later. Priming information ed. Arie W. Kruglanski and E. Tory Higgins, NY: Guilford,
search procedures might also be enduring. 201-31.
Finally, it is important to note that although the activation Herr, Paul M. (1986), "Consequences of Priming: Judgment and
and use of procedural knowledge can often be influenced Behavior," Joumal of Personality and Social Psychology, 51
by situational factors, procedures that become very well (6), 1106-15,
leamed through repeated application may become chroni- Herr, Paul M., Frank R. Kardes, and John Kim (1991), "Effects
cally accessible in memory and be applied independently of Word-of-Mouth and Product-Attribute Information on Per-
of context. Under these conditions, priming effects of the suasion: An Accessibility-Diagnosticity Perspective," Joumal
of Consumer Research, 17 (March), 454-63.
sort we identified in the present research may have less
Higgins, E. Tory (1997), "Beyond Pleasure and Pain," American
impact. Psychologist, 52 (December), 1280-1300,
Higgins, E. Tory, William S. Rholes, and Carl R. Jones (1977),
"Category Accessibility and Impression Formation," Joumal
of Experimental Social Psychology, 13 (March), 141-54.
REEERENCES Houston, David A., Steven J. Sherman, and Sara M. Baker (1991),
Adaval, Rashmi and Kent B. Monroe (2002), "Automatic Con- "Feature Matching, Unique Features, and the Dynamics of
struction and Use of Contextual Information for Product and the Choice Process: Predecision Conflict and Postdecision
Price Evaluations," Joumal of Consumer Research, 28 Satisfaction," Journal of Experimental Social Psychology, 27
(March), 572-88, (September), 411-30.
Anderson, John R, (1983), The Architecture of Cognition, Cam- Huber, Joel, John W. Payne, and Christopher Puto (1982), "Adding
bridge, MA: Harvard University Press, Asymmetrically Dominated Altematives: Violations of Reg-
PROCEDURAL PRIMING 737

ularity and the Similarity Hypothesis," Joumal of Consumer Simonson, Itamar (1989), "Choice Based on Reasons: The Case
Research, 9 (June), 90-98. of Attraction and Compromise Effects," Joumal of Consumer
Kirmani, Amna, Michelle P. Lee, and Carolyn Yoon (2004), "Pro- Research, 16 (September), 158-74,
cedural Priming Effects on Spontaneous Inference Forma- Skowronski, John J. and Donal E, Carlston (1989), "Negativity
tion," Journal of Economic Psychology, 25 (December), and Extremity Biases in Impression Formation: A Review of
859-75.
Explanations,"/'.s>'c/!o/ogica/BM//erm, 105 (1), 131-42,
Klein, Jill G, and Rohini Ahluwalia (2005), "Negativity in the
Smith, Eliot R, (1990), "Content and Process Specificity in the
Evaluation of Political Candidates," Joumal of Marketing, 69
(January), 131^2. Effects of Prior Experiences," in Advances in Social Cogni-
Kolers, Paul A. (1976), "Reading a Year Later," Joumal of Ex- tion, Vol, 3, ed, Thomas K, Srull and Robert S, Wyer, Hills-
perimental Psychology: Human Learning and Memory, 2 dale, NJ: Erlbaum, 1-59.
(September), 554-65. (1994), "Procedural Knowledge and Processing Strategies
Ofir, Chezy and Itamar Simonson (2001), "In Search of Negative in Social Cognition," in Handbook of Social Cognition, 2nd
Customer Feedback: The Effect of Expecting to Evaluate on ed,, Vol, 1, ed, Robert S, Wyer and Thomas K, Srull, Hillsdale,
Satisfaction Evaluations," Joumal of Marketing Research, 38 NJ: Erlbaum, 101-51,
(May), 170-82. Smith, Eliot R,, Nyla R, Branscomhe, and Carol Bormann (1988),
Ostrom, Thomas M. and Harry S. Upshaw (1968), "Psychological "Generality of the Effects of Practice on Social Judgment
Perspective and Attitude Change," in Psychological Foun- Tasks," Journal of Personality and Social Psychology, 54 (3),
dations of Attitudes, ed. Anthony G. Greenwald et al.. New 385-95,
York: Academic Press, 217-42. Tversky, Amos and Daniel Kahneman (1974), "Judgment under
Park, Jong W. and Jung K. Kim (2005), "The Effects of Decoys Uncertainty: Heuristics and Biases," Science, 185 (Septem-
on Preference Shifts: The Role of Attractiveness and Provid- ber), 1124-30,
ing Justification," Joumal of Consumer Psychology, 15 (2),
Wyer, Robert S. (2004), Social Comprehension and Judgment: The
94-107.
Role of Situation Models, Narratives, and Implicit Theories,
Schneider, Walter and Richard M. Shiffrin (1977), "Controlled and
Automatic Human Information Processing. I. Detection, Mahwah, NJ: Erlhaum,
Search, and Attention," Psychological Review, 84 (1), 1-66. (2008), "The Role of Knowledge Accessibility in Cog-
Schwarz, Norbert and Robert S. Wyer (1985), "Effects of Rank nition and Behavior: Implications for Consumer Information
Ordering Stimuli on Magnitude Ratings of These and Other Processing," in Handbook of Consumer Research, ed, Curtis
Stimuli," Joumal of Experimental Social Psychology, 21 (1), Haugtvedt, Frank Kardes, and Paul Herr, Mahwah, NJ: Erl-
30-46. baum, 31-75,
Shafir, Eldar (1993), "Choosing vs. Rejecting: Why Some Options Xu, Jing A, and Robert S, Wyer (2008), "The Effect of Mindsets
Are Both Better and Worse than Others," Memory and Cog- on Consumer Decision Strategies," Joumal of Consumer Re-
nition, 21 (4), 546-56, search, 35 (February), forthcoming.