Content uploaded by Priya Raghubir
Author content
All content in this area was uploaded by Priya Raghubir
Content may be subject to copyright.
Position-based Beliefs: The Centre-Stage Effect
May 27, 2015
1
Position-based Beliefs: The Centre-Stage Effect
ABSTRACT
This paper examines the existence and consequences of consumers’ position-based beliefs
about product layouts. We propose that consumers believe that options placed in the center of a
simultaneously presented array are usually the most popular. This belief translates into their
choosing options placed in the centre more often than those on the sides of a display: the
Centre-Stage Effect (Studies 1 and 6). Results are driven by inferences of product popularity
rather than higher levels of attention to products in a given position (Studies 2 and 3). The
preference for middle options is attenuated when layout-based information is not diagnostic
(Study 5), and accentuated when people explicitly take into account other people’s preferences,
increasing the need to choose a popular option (Study 5). Increasing the accessibility of own
preferences for the intrinsic attributes about the products reduces the use of position-based
beliefs to make judgments and attenuates the centre-stage effect (Studies 5 and 6). Theoretical
implications for marketplace meta-cognitions, position effects, the use of overall beliefs versus
individuating information to make judgments, and visual information processing are discussed.
2
This paper examines the existence and consequences of consumers’ position-based
beliefs about product layouts. Position-based beliefs belong to the larger genre of marketplace
meta-cognition that refers to individuals’ thinking about the rules that govern marketplace
interactions (Wright, 2002). Marketplace meta-cognitions include consumers’ implicit theories
about how companies exert marketplace persuasion and influence, that is, consumers’ sets of
beliefs about marketers’ general tactics in typical marketplace interactions (Friestad & Wright,
1994). We propose that consumers hold beliefs about the rules that marketers follow in their
physical ordering of products within arrays and, accordingly, use the position of a product in an
array as a source of information to construct product judgments, especially in the absence of
alternative sources of information.
Overall, there are certain position rules that seem to govern the physical ordering of
people, items, and things across contexts and domains. In a print advertisement, consumers are
used to seeing brand name information on the right and logos on the left (Janiszewski, 1990).
Within a store, consumers expect that products located in an end-of-aisle display to be on
discount (Inman, McAlister & Hoyer, 1990). In a group task, observers believe that people
seated in the centre are the most important (Raghubir & Valenzuela, 2006) and influential
(Taylor & Fiske, 1975). Given that research has found that people assign meaning to the position
of an item in an array, in the absence of alternative information, an item’s position could be used
as a source of information and affect consumers’ preferences and choice. However, prior
research examining the effect of the physical position of products in an array has found
inconsistent effects ranging from preferences for the last option (Nisbett & Wilson, 1978), the
middle option (Christenfeld, 1995; Shaw, 2000), either the first or the middle option (Dreze, Hoch
& Purk, 1994), or no position effect at all (Chandon et al., 2007). Prior research is also divided as
to why position effects occur. While attention has been proposed to be the mediating route
driving effects of position on choice (Nisbett & Wilson, 1978), evidence regarding this route is
lacking, both using indirect measures of recall (Shaw, 2000; Taylor & Fiske, 1975), as well as
3
direct measures of eye-tracking (Chandon et al., 2007). This has led researchers to argue that
preferences for a specific position reflect a simple rule that people apply to minimize mental
effort (Raghubir & Valenzuela, 2006) outside of their consciousness (Christenfeld, 1995).
This paper develops an understanding of when products placed in the center of an array
will enjoy an advantage -- the “Centre-Stage Effect,” while at the same time understanding why
they do so. We investigate the specific nature of the belief that consumers have about products
placed in the centre, and establish whether, when, and why this belief affects their judgments
and choices. Specifically, we propose that consumers believe that marketers organize choice
sets so as to represent consumers’ preference structure, with the modal or most liked market-
place option being that in the middle. As a consequence, consumers infer that a product, which
is placed in the middle of an array, has a high chance of being the most popular. A popular
product is defined as one that is the most preferred and chosen by other consumers, and has
the highest market share in its category. When consumers’ goals encourages the use of the
“consensus heuristic,” or the general judgment rule that other peoples’ opinions are accurate
(Chaiken, 1987), then popularity inferences of the product placed in the middle of an array
translate into favorable attitude judgments and, thus, product preference (Chaiken &
Maheswaran, 1994). We test our proposed belief-based model against the alternative attention-
based model (where products placed in the centre are more visually salient, attract more
attention than others, and are, therefore, more likely to be chosen), and find support for the
former.
Specifically, Studies 1-3 demonstrate that consumers i) hold shared beliefs that the most
advantageous location for a product in an array is the central position, ii) believe that products
placed in the middle are the most popular, iii) evaluate products in the centre more favorably,
and iv) choose products in the centre more often than chance levels. Studies 4-6 show that the
preference for central positions are stronger when v) the contextual cue of position is perceived
to be informative (versus uninformative) of market-place tactics, vi) consumers consider the
4
preferences of others (versus themselves) in their decision, and vii) attitudes towards the
products to be judged are less (versus more) accessible.
In the following sections, we describe the extant literature on position effects and review
the evidence for the two routes proposed for position effects: attention and belief-based
inferences. We then develop a set of hypotheses proposing the mediating effect of popularity
perceptions on preferences, and test our predictions. The general discussion concludes with
theoretical implications about the route through which position effects occur: lower-order
attention mechanisms versus higher-order cognitive beliefs. As this paper examines arrays of
products that are presented simultaneously (e.g., product arrays that consumers may encounter
when shopping online1), we discuss managerial implications for product managers and online
retailers. We conclude with study limitations and areas for future research.
POSITION EFFECTS: Prior Literature
Which Position has an Advantage?
A position advantage is defined as a more favorable evaluation, a higher choice
likelihood and/ or higher sales of a product occupying a specific position in an array. Extant
findings are inconsistent; supporting an advantage for the last option presented (Nisbett &
Wilson, 1978), for items at the end of a display (Inman et al., 1990), for items placed in the
middle (Christensen, 1995; Shaw, 2000), for both extreme and/ or middle positions contingent on
the product category (Dreze, Hoch & Purk, 1994), as well as no position advantage (Chandon et
al., 2007). This literature is briefly reviewed below.
In the classic study by Nisbett and Wilson (1978), participants shown a series of
stockings sequentially were more likely than chance odds to choose the last option with which
1 Examples include www.Amazon.com or www. TravelSmith.com. Online product displays are an
appropriate context in which to understand the current research as online stores typically allocate one
shelf “facing” to each product, and, thus, make shelf space allocation non-relevant and also facilitate visual
processing by providing small and easily examinable electronic shelves, which one encounters head on
(Breugelmans, Campo & Gijsbrechts, 2007).
5
they were presented. In a different context, Inman, McAlister and Hoyer (1990) found that
products placed at the end of the aisle of a display enjoyed higher sales even when they were
not offered at a price cut, also suggesting an advantage for extreme positions.
On the other hand, Shaw (2000) found a preference for the central position using
graphical stimuli within a three-poster collage. Shaw’s results are consistent with those of
Christenfeld (1995) who examined choice behavior among arrays of identical items. He found
that individuals reliably chose items located in the middle position when picking products from
supermarket shelves, choosing bathroom stalls, choosing a toilet paper roll within a stall, and
picking one among a row of arbitrary symbols.
Both a central-position and an extreme-position advantage were documented by Drèze et
al. (1994) contingent on product category. They conducted a field study to show that moving a
product from the worst to the best horizontal movement increased sales by as much as 15%,
though the best positions were either those in the middle or those on the sides. Finally, Chandon
et al. (2007) examined choice behavior using simulated grocery shelves and did not find
consistent effects of position on choice.
One way of reconciling the inconsistent findings regarding which position affords an
advantage is by examining the differences in the procedures and contexts of the different
studies. One difference between Nisbett and Wilson’s (1978) finding that the last position is
preferred and Christensen’s (1995) finding that the middle position is preferred, is that options
were presented sequentially in the former and simultaneously in the latter. However, in the
Inman et al. (1990) study as well as the Dreze et al. (1994) and the Chandon et al. (2007)
studies, options were presented simultaneously, but while Inman et al (1990) found a preference
for the extreme (end-of-aisle) position, Dreze et al (1994) found preferences for both central and
extreme positions, and Chandon et al. (2007) found no reliable position effects on choice. Thus,
sequential versus simultaneous presentation, while a potential moderator of the effect, is not a
sufficiently parsimonious explanation for inconsistent position effects. Having said that, when
6
options are presented sequentially, findings seem to support either primacy or recency–based
explanations (Schuman & Presser, 1996), suggesting that simultaneous presentation is more
likely to foster a middle position advantage. As our focus is to identify why people prefer the
central position, we employ a simultaneous presentation mode in our experiments.
Prior research on position effects also differ in terms of the explanation provided for
these. Explanations are categorized as attention-related (Route 1) and belief-related (Route 2)
and are discussed below.
Route 1 for Position Effects: Position Salience Attention Preference
Salience of stimuli is defined as the aspect of a stimulus that makes it stand apart from
other stimuli due to its inherent characteristics or its context. Commonly studied antecedents of
salience include the movement, brightness, presence of warm-colors, complexity, and proximity
of an object to the perceiver (McArthur & Post, 1977). Salience of an object makes it more likely
to be attended to, recalled, and used to form attitudes in an information aggregation task (for a
review on the antecedents and consequences of salience effects, see Fiske & Taylor, 1989).
Given the literature on salience effects, it is plausible that an object that is at the centre of
an array will be more salient, receive more attention, and, for this reason, be preferred. In fact,
Nisbett and Wilson (1978) explained their results for the stockings study in terms of higher
attention paid to the last option presented in a sequence (Schuman & Presser, 1996). However,
research examining the antecedent role of attention for position effects using simultaneous
presentation of stimuli has found little or no evidence for it. The earliest exploration was in a
person perception domain using a recall task where the person in the middle, though found to be
more influential in a group task, was not recalled more than others (Taylor & Fiske, 1975).
In the domain of product perception, Shaw (2000) ruled out an attention-driven
mechanism driving the preference for the central position by showing no difference in the
number of graphic items recalled from posters in the left, centre and right position within a three-
poster collage. He, therefore, discarded a focus of attention based explanation of centrality
7
preferences. However, it is plausible that the indirect test of recall is not powerful enough to
diagnose whether attention has been differentially paid to stimuli at the time of making a choice.
Given the limitation of the recall task, recent research has directly tested the mechanism
of attention using the eye-tracking methodology (Chandon et al., 2007). Chandon et al. (2007)
found no evidence that higher attention to specific position led to more favorable evaluations and
choices at higher than chance likelihood. They concluded that though the center of a display
may be more likely to be noticed, the resulting visual lift did not carry through to product
consideration and choice, which did not show a robust position effect.
To summarize, empirical support is lacking for attention being an antecedent for centrality
preferences when stimuli is presented simultaneously. This is not surprising as research on
vision effects using simultaneous presentation predicts that the item left of centre (versus the
centre) is most noticeable (Ducrot & Pynte, 2002). Consumer psychologists have also shown
that items on the right or left of a display are preferred contingent on whether they are visual or
verbal stimuli (Janiszewski, 1990). Given the weak theory and evidence for an attention-driven
mechanism explaining position effects for simultaneous presentation, we next examine the role
of cognitive beliefs driving preference and choices for a given position.
Route 2 for Position Effects: Beliefs Position Inferences
Four distinct streams of literature, in contexts ranging from people perception and survey
methods to product perception, all suggest that people draw inferences from position, and that
these inferences involve perceptions that the option in the middle is the most common in the
population (i.e., the most popular). These are discussed below and used to reconcile seemingly
inconsistent position effects reviewed earlier.
People Perception. The social psychology literature has shown that there are certain
beliefs associated with the meaning of the position people occupy in a group setting (McArthur &
Post, 1977; Raghubir & Valenzuela, 2006; Taylor & Fiske, 1975). Taylor and Fiske (1975) were
8
the first to examine the effect of visual placement on evaluations of people’s performance. They
found that people placed in middle positions, facing onlookers, were perceived to be more
causal to a group’s outcome and concluded that this was due to the salience of being in the
central position. McArthur and Post (1977) speculated that this position effect could be due to
cultural norms and beliefs: In the real world people facing an audience are more prominent than
those with their backs to an audience and those who sit at the centre of a table are typically the
most important individuals at the table.
Following up on McArthur and Post’s (1977) speculation, Raghubir and Valenzuela
(2006) showed that players assigned at random to one of eight positions in a semi-circular
horizontal array in the television show “The Weakest Link,” were more likely to win the game
when they started off in one of the two central positions. Raghubir and Valenzuela (2006) argued
that this was due to people’s beliefs, based on learned associations, that important people are
expected to sit in the middle (e.g., the CEO in a group interviewer panel). If people believe that
those in the centre are better than those in non-central positions then they may substitute this
belief instead of spending the resources to process individuating information about each of the
candidates (Fiske & Neuberg, 1990). Therefore, prior beliefs would dictate how attention was
allocated to the different positions. Follow-up studies demonstrated that the position advantage
for those in the centre was, indeed, due to their errors being paid less attention to than the errors
of others, especially by study participants who held the belief that “important people sit in the
middle.” Thus, counter-intuitively, it was the lack of attention paid to the individuating information
about people in the center that led to their enjoying an advantage: they were “Center[s] of
Inattention” (Raghubir & Valenzuela, 2006).
Translating these people perception findings that centrality implies prominence, or
importance, to a product domain; we suggest that items in the middle may have the best
reputation or standing. It could imply either beliefs regarding higher product quality or overall
market power (i.e., popularity). Thus, while the literature in people perception suggests that
9
central positions may have an advantage, it does not predict the specific nature of the beliefs
people have (and the types of inferences they would draw) about products placed in the centre.
The literature on survey methods assists in specifying the nature of centre-related beliefs. This is
discussed next.
Survey Methods. The literature on the cognitive aspects of survey methodology can help
understand the specific rule governing what option people choose, and why, when options are
simultaneously presented. This literature has documented that responses to multiple-choice
questions are frequently constructed on the basis of the contextual cues available in the
questionnaire rather than being retrieved from memory (for a review see Sudman, Bradburn &
Schwarz, 1986). This process of identifying an appropriate option is conceptually similar to
making a choice between alternatives presented in an array. In one of the first demonstrations of
how the range of response alternatives presented could affect the actual responses elicited,
Schwarz et al. (1985) showed that response alternatives presented using a low range of
frequencies (e.g., < ½ hour, ½-1 hour, etc.) elicited a lower response than those elicited using a
high range of frequencies (< 2 ½ hours, 2 ½-3 hours, etc.). They further showed that people
exposed to the high range of response alternatives inferred that the average frequency of
performing the given behavior in the population was higher than those exposed to the low range
of response alternatives. This was presumably because the respondents believed that the
researcher had designed the choice set in a manner reflecting the distribution of frequencies in
the population, an idea consistent with Grice’s (1975) theory of “conversation norms.” Schwarz
et al. discussed their findings in terms of a consumer process whereby consumer assess what
the population average is given the range of response alternatives, and then judge whether they
perform the behavior at a frequency lower than, equal to, or higher than the population average.
Most pertinent to the current investigation is the idea that people believe that a response set is
constructed such that the middle option reflects the population average.
10
Taking the idea further, Menon, Raghubir and Schwarz (1995) showed that even when
the set of response categories do not affect reports of behavioral frequencies (e.g., because
people are able to retrieve these from memory and do not need to construct them on the basis of
contextual cues), respondents continue to expect the middle option to reflect the population
average. Said differently, the inference from the response categories appears to be a robust
effect, although it is not always reflected in choice. The inference from the range of response
alternatives appears to be strong enough that it can affect responses to later questions (related
to the behavioral frequency) even when it does not affect responses to the behavioral frequency
itself (Menon, Raghubir & Schwarz, 1997). Thus, in a survey methods context, respondents infer
that options in the middle are the most common in the population. Applied to the domain of
product placement, this would imply that middle options are perceived to be the modal choice of
consumers, or the most popular options. The literature in behavioral decision-making also
supports this idea.
Behavioral Decision Theory. In any given choice set, the literature in behavioral
decision-making shows that: i) consumers believe that a given selection of alternatives reflects
the population range (Wernerfelt, 1995), with preferences normally distributed along this range
(Coombs & Avrunin, 1977); ii) consumers prefer the compromise option which is conceptually in
the middle of a choice set to extreme options (Simonson & Tversky, 1992). Therefore, in both a
survey methods context, as well as in a choice context where choices are presented
simultaneously, prior research has demonstrated that people believe that the most popular
option is in the middle of the set. These sets are typically ordered in decreasing or increasing
order based on a variable, rather than representing nominally different options. Even when
response categories are nominal, however, as in the case of multiple choice questions on a quiz,
test takers have a strong and systematic tendency to hide correct answers in middle positions
when they would like to make the test difficult, presumably because they believe that other test
takers will presume that test-makers have not used the most likely option for the correct answer
11
(Attali & Bar-Hillel, 2003). Therefore, it is plausible that even when options are nominal (rather
than ordinal), people believe that the option chosen by the most people will still be the option in
the center of an array. Specifically, people believe that central options are the most popular.
Marketplace Meta-cognition. Prior literature in marketing has shown that many judgments
are based on consumers’ beliefs about the functioning of a market-place: consumers have
theories about how the level of advertising expenditure relates to brand importance (Kirmani &
Wright, 1998), how different marketing mix decisions (e.g., channel decisions) relate to brand
positioning (Pham & Johar, 1997) or which type of brand is usually given precedence in
marketing communications (Buchanan, Simmons & Bickart, 1999). These are all examples
within the larger genre of marketplace meta-cognitions (Wright, 2002). Given the ubiquity of
arrays in consumers’ choice sets, it is plausible that consumers have developed theories about
which types of products are placed in different positions in an array: specifically, we propose that
one of those beliefs is that popular products will be found in the center.
In the domain of product placement, in view of the fact that the prominence of a particular
position did not seem to fully explain position advantages, Christensen (1995) proposed that the
central position advantage in choice could reflect a simple rule that people unconsciously apply
to minimize mental effort. This proposition is consistent with the large and growing evidence for
marketplace meta-cognition (Buchanan, Simmons & Bickart, 1999; Friestad & Wright, 1994;
Kirmani & Wright, 1998; Pham & Johar, 1997; Wright, 2002). If consumers have shared beliefs
that marketers place products with shared characteristics in specific positions, then they may
use the fact that a product occupies a particular position to make inferences about its
characteristics.
In fact, Inman et al. (1990) reasoned that the reason for the sales increase for a product
on display at the end of the aisle was that consumers inferred that it had a price cut based on
their beliefs that displays are usually associated with price cuts. In this sense, Inman et al’s
(1990) findings and reasoning are consistent with those of Christenfeld (1995), despite the fact
12
that Inman et al. found a position advantage for extreme positions, and Christensen found it for
central positions. It is plausible that consumers have different rules for products placed in
different positions in an array, leading to different positions enjoying an advantage depending on
consumers’ goals and product category characteristics. Therefore, a cognitive-belief based
antecedent for position effects can also help understand Dreze et al.’s (1994) results where
some products had increased sales when they were placed on the extremes of a display and
others had increased sales when they were in the centre.
To summarize, researchers have i) not found a consistent preference for a specific
position in an array; ii) not found that attention mediates preference formation; iii) found
evidence that people have beliefs reflecting the meaning that different positions hold, although,
this could lead to an advantage for extreme or middle positions. Based on this literature, we
investigate the specific nature of the rule governing choice of products in central positions. Our
reasoning and hypotheses are presented next.
The Centre-Stage Effect
If consumers infer that products placed in the middle of an array are those chosen by the
majority of other consumers and, therefore, popular, then in the absence of alternate memory- or
context-based information regarding the alternatives, they could use the inference of popularity
as a cue to construct their own preferences (Feldman & Lynch, 1988). In fact, attitude research
has shown that the “consensus heuristic,” or the belief that a product is good if other people like
it, is a strong heuristic that affects attitudes when people have low motivation or ability to
evaluate individuating information about an attitude object (Chaiken & Maheswaran, 1994).
This suggests:
H1: Consumers prefer an option placed in the centre to options placed at either extreme of a
set of nominally ordered alternatives presented simultaneously.
H2: Consumers believe that options placed in the centre are more popular than those placed
at either extreme of a set of nominally ordered alternatives presented simultaneously.
13
H3: Consumers beliefs regarding popularity mediate their preferences for options placed in
the centre.
Study 1 test for the existence of a centre-stage advantage (H1). Study 2 attempts to
replicate H1 using a different product category, set size, and experimental procedure. It also
investigates whether this advantage can be explained by higher levels of attention paid to the
middle option. Study 3 tests H1, H2, and H3. It examines whether the centre-stage advantage is
due to inferences regarding a product’s popularity that translate into favorable attitudes
regarding the product while at the same time attempting to replicate Study 2 results by rigorously
ruling out the alternative explanation that the centre-stage effect is due to higher attention paid to
the middle option.
Factors Moderating the Centre-Stage Effect
A belief-based inference mechanism also suggests that position effects should be
contingent on the use of alternate sources of information to make a judgment. One such source
of information is the intrinsic appeal of the product itself. Given that our focus is to understand
when and why people prefer the central position to other positions, in our empirical approach we
hold brands and prices constant across options to minimize any brand/ price-related inferences
that consumers may make. The idea that a position advantage is not ubiquitous, but contingent,
allows for the reconciliation of Chandon et al.’s (2007) results with those of Inman et al. (1990),
Shaw (2000) and Christensen (1995). It is possible that the use of eye-tracking mechanisms led
to greater attention being paid to stimuli which attenuated the use of belief-based inferences,
leading to null effects of position on sales.
The overall theory we have put forward is that consumers choose a product in the centre
because they use the overall belief of the middle product being the most liked by others (most
popular) as a substitute for actual product information. Prior research has shown that when pre-
existing attitudes are available, relevant and easily retrieved, then, attitudes are less likely to be
14
constructed on the spot on the basis of contextual cues (Feldman & Lynch, 1988; Menon et al.,
1995; Simonson & Tversky, 1992). When attitudes are relevant and accessible, consumers can
use these attitudes instead of constructing them based on contextual cues, such as spatial
position. If attitudes are not applicable or inaccessible and need to be constructed, then
consumers are more likely to use the salient spatial cue of position, leading them to infer that
products in the centre are better.
Attitude relevance depends on the decision context and on consumers’ assessment of
what type of judgment is suitable to make the decision at hand (Snyder, 1982; Fazio, 1990). For
example, if the importance of the preference of others becomes salient in the decision process,
contextual spatial cues may be used to assess product popularity. However, as the use of a
source of information to make judgments is inversely related to the accessibility and diagnosticity
of alternate sources of information that can be used to make the same judgment (Feldman &
Lynch, 1988), we predict that the use of contextual spatial order to make product inferences will
be attenuated when attitudes about the products are less accessible, and/ or when the belief is
not perceived to be diagnostic. This implies that the effect is contingent and should be more
likely to be found when: i) the context encourages inferences of popularity, ii) choosing popular
items is consistent with purchase goals, and iii) when attitudes based on the intrinsic product
information are less accessible. Formally:
H4: The preference for items in the centre is greater the more informative the order-of-
presentation is believed to be.
H5: The preference for items in the centre is greater the more people consider the
preferences for others (versus themselves).
H6: The preference for items in the centre is greater when choices are elicited prior to (versus
subsequent to) individual preferences.
Studies 4, 5 and 6 test H4-H6.
15
POSITION EFFECTS: Summary of Empirical Approach
We conducted one pretest and six studies to examine H1-H6. Studies 1 and 2 test for the
center-stage effect (H1) using different contexts, procedures, set sizes and measures. Study 3
examines the mediating path behind the center-stage effect (H2 and H3). Studies 4-6 examine
the moderating conditions surrounding the center-stage effect (H4-H6), with study 5 using a
moderated mediation approach to develop further support for the antecedents of when and why
people prefer products in the center. In each study we attempt to replicate prior effects before
testing for process explanations via mediation paths, moderation, and moderated mediation
analyses. Similarities across study participants, design, measures, and analyses are
summarized in the “Overall methodology” section below, with individual differences in the
method and manipulations for each study described in their “Method” section.
Overall Methodology
Study Participants. Experimental participants in all studies, except Study 3, were
students enrolled in an introductory business class who participated for partial course credit
(Pretest = 188, Study 1 = 48, Study 2 = 62, Study 4 = 44, Study 5 = 151, and Study 6 = 180).
There was no overlap in participants across studies. In Study 3, participants were recruited
throughout the campus through a behavioral laboratory (n = 136).
Design and Procedure. The size of the choice sets varied across studies: three options
(Studies 1 and 6), five options (Studies 2, 3, and 5), and nine options (Study 4). In all studies
(except Study 3), the order in which a set of stimuli was shown was manipulated across
conditions, such that every option was presented in each of the available positions.
In Studies 1 and 6, participants were told that as a token of appreciation for their
participation they were going to receive a packet of chewing gum. They could choose from three
varieties: Spearmint (S), Peppermint (P) and Winterfrost (W) that were presented in three
different orders with their choice recorded.
16
In Studies 2-5, participants were told to imagine the following scenario:
“You are shopping for a BYO (bring your own drink) party where guests will bring their own
drinks, but you need to pick up snacks and starters for everyone. Cost is not an issue as
everyone is going to equally share the overall costs of the party. You simply need to make a
choice of one item.”
In all studies (except Study 4 where the choice set was nine varieties), participants were
shown five different pretzel varieties (butter checkers, tiny twists, cheddar cheese, honey
mustard, and braided twists) of the same brand (Rold Gold) and told that they were all priced at
$2.19 (to control for any price inference effects; Inman et al., 1990). In all studies (except study
3) the varieties were presented in different orders so that each variety occupied each of the
different positions across conditions. Subsequent to being exposed to the stimuli, participants
responded to a pencil and paper questionnaire.
Measures. Measures common across studies are described below.
1. CHOICE-RELATED MEASURES. Study 1 unobtrusively recorded choice from among three
items, studies 2-5 asked participants to rank-order their preferences, and study 6
explicitly asked respondents to state their choice. Studies 2, 3, and 5 measured
purchase intentions for all items presented (“1=Definitely will not purchase” and
“7=Definitely will purchase”), with Studies 3 and 5 additionally measuring liking (“1=Not
at all,” and “7= Very Much”).
2. POPULARITY INFERENCE-RELATED MEASURES. Studies 2-6 asked participants to rate the
popularity of each item (“1=Not at all popular,” and “7=Very popular”). Studies 3 and 5
also elicited estimates of relative market share, and estimates of what percentage of
space should be allocated to each item (total = 100% for both). Participants were told
“Assume you are managing a store and need to stock a range of products that
are commonly purchased by students. You need to allocate shelf space on the
basis of each product’s popularity. That is, items that are likely to be chosen by
the most people should get more space allocation than items that are chosen by
only a few people. Decide what percentage of your shelf space you will allocate
to each variety (_%).
17
An exploratory factor analysis across the three popularity measures revealed a five factor
solution in both studies (Variance explained = 85.05% and 81.09% in Studies 3 and 5
respectively, factor loadings > .75), such that the measure of share allocation, rating of
popularity, and the estimate of market share for each variety loaded onto a single factor for each
of the five positions. Therefore, the three measures were aggregated to form a single popularity
index for each position. The reliability of the five indices was acceptable (Study 3 αs = .75, .76, .
72, .73 and .74; Study 5 αs = .68. .71, .66, .74, and .70 for positions 1-5 respectively).
3. ATTENTION-RELATED MEASURES. Studies 2, 4 and 5 measured unaided recall using a
surprise recall task administered at the end of the questionnaire as a surrogate for
attention (Taylor and Fiske 1975). Participants were asked to write down the names of
as many varieties of the stimuli to which they were exposed to as they could remember.
Studies 3 and 5 used a different set of measures described in their method section.
4. EXPERIMENTAL VALIDITY MEASURES. In all studies where self-reports of judgments were
elicited (Studies 2-6) we assessed study participant’s level of motivation while answering
the questionnaire, as well as their judgments of how realistic the scenario was (1 = Not
at all/ 7 = Very) at the end of the questionnaire. For all studies the means were at or
above the scale midpoint for both motivation (M = 4.95, 4.98, 4.52, 3.87 and 4.85, for
studies 2-6 respectively), and task realism (M = 5.24, 5.24, 5.20, 4.21, and 4.65, for
studies 2-6 respectively).
Analyses. Analyses are in terms of the position (e.g., 1st - 5th) that the variety occupies in
the array incorporating other experimental manipulations as a between subjects factor.
1. For categorical variables (choice, rank, recall likelihood), we tabulate responses across
experimental conditions, report an overall χ2 and cell percentages.
2. For continuous variables (interval scaled judgments, open-ended estimates, rank in
Study 3, recall position, and vividness of image recalled), we conduct a repeated
18
measures ANOVA, where the position factor is the repeated measure. We report the
quadratic contrast for position which captures the inverse U-shaped function predicted
by H1 and H2: evaluations are more favorable in the centre than at the extremes.
3. Mediation tests use the procedure recommended by Baron and Kenny (1986):
a. Demonstrating the effect of the independent variable on the dependent variable.
b. Demonstrating the effect of the independent variable on the proposed mediator.
c. Including the proposed mediator as a predictor in the first analysis.
If the covariate is significant while the effect of the hitherto significant independent variable on
the dependent variable is no longer significant (perfect mediation) or reduced (partial mediation),
this supports mediation. Study 3 uses ANOVA for tests (a) and (b) and ANCOVA for (c) to test
mediation. Study 5 tests for moderated mediation by separating the two conditions where
different mediation paths are expected, and conducting mediation analyses using ANOVA and
ANCOVA for each condition separately. The studies are now described.
Pretest
The goal of the pre-test was to assess whether consumers believe that a certain position
in an array is advantageous.
Method. Participants were told:
“You are a product manager and you want consumers to choose your product X, more
than the competitors’ products A, B, C, and D. If you had a choice to decide your shelf position,
where would you ideally like to have your product X placed.”
Five empty boxes placed from left to right were provided.
Results. A majority of participants chose the middle position (63.44%, 118/ 186) over the
other positions (Extreme left = 35, Left of centre = 18/ 186; Right of centre = 9/ 186; Extreme
Right = 9/186; χ2 > 300, p< .001) supporting the argument that consumers believe that the
middle position is the most advantageous.
19
Study 1: The Centre-Stage Effect
The goal of this study was to test the center-stage effect (H1): the prediction that people
prefer the middle option in an array to options at the extremes of the array.
Method. Participants’ choice of one of three chewing gums, ostensibly provided to thank
participants for their participation in an experiment, was unobtrusively recorded.
Results. The chewing gum variety placed in the middle was chosen half the time (24/48),
while the one on the left was chosen 29.17% of the time (14/48) and the one on the right chosen
20.83% of the time (10/48;
χ
2 = 6.50, p < .05). Thus, choices favor the middle position,
supporting the “centre-stage” effect, and replicating Christenfeld’s (1995) results.
Study 2: Assessing Attention as a Mediator of the Centre-Stage Effect
The goal of this study was to replicate study 1 results using a different procedure, and
examine whether it can be explained in terms of greater attention paid to the middle item.
Method. We increased the size of the choice set from three to five, changed the product
category from chewing gum to pretzels, measured preference using a rank task (vs. choice), and
also measured unaided recall using a surprise task as a surrogate for attention.
Results. Participants ranked the variety in the centre as their first preference 30% of the
time (vs. 14%, 22%, 23%, and 11% for varieties in the 1st, 2nd, 4th and 5th positions respectively,
χ
24 = 7.00, p < 0.05). There were no differences in the percentage of times a variety was recalled
as a function of its position (59.7%, 62.1%, 53.4%, 53.4% and 51.7% for 1st – 5th position, n.s.).
In other words, although the variety in the first position is recalled as often as the variety in the
middle position, it is chosen significantly less often. As a consequence, the difference in choice
behavior by position cannot be explained by differences in recall. These results replicate those of
Shaw (2000). They suggest that preferences for central positions cannot be explained in terms
20
of higher attention paid to those positions, using recall as a proxy for the level of attention paid.
We now examine whether popularity inferences could explain the centre-stage advantage.
Study 3: The Mediating Role of Popularity Inferences on the Centre-Stage Effect
The goal of this study was to examine whether consumers infer that the middle option is
more popular and whether these inferences mediate preferences for the middle option (H2 and
H3). At the same time, we wished to increase the evidence that inferences, rather than attention
mediate the route to preference for the middle option, by replicating Study 2 results using direct
as well as indirect measures of attention.
Method. In this study the set of dependent measures was increased. Given that the order
in which measures are elicited can lead to carry-over effects to other measures, which would
makemediation difficult to establish, we fully counterbalanced the order of three groups of
dependent variables (preference, attention, and popularity inferences) across study conditions,
resulting in six different orders of elicitation of measures. The order of elicitation is included as a
between-subjects factor in the five-level (position) repeated measures ANOVAs for each of the
variables below. The main or interaction effect of order is not significant unless mentioned.
Further, to reduce consistency pressures, the study was presented as a series of short
unrelated studies that were being conducted as part of pilot research. Each set of measures was
titled as a separate study, and used a different font to make it look visually different. The choice
measures were titled “Shopping for a Party,” the attention measures were titled “Packaging
Decisions,” and the popularity measures were embedded in a “Store Management” study. There
were unrelated filler tasks administered between the three sets of measures to erase short term
memory and reduce any consistency pressures and carry-over order effects.
We used a single order of presentation for the five pretzel varieties as our goal was to
examine mediating paths between dependent variables rather than to establish the
generalizability of the “centre-stage” effect. The measures are described below:
21
1. PREFERENCE was elicited using three measures: rank-order (lower numbers reflect more
favorable evaluations), purchase intentions, and liking (both elicited on 7-point scales
with higher numbers reflecting more favorable evaluations). An exploratory factor
analysis across the three evaluation measures revealed a five factor solution (84.71%
variance explained), such that the measure of rank (negative sign), intention and liking
for each variety loaded onto a single factor for each of the five positions. However, as
intention and liking were not elicited for Study 2, to assess whether results replicate
using the same measure, we analyze the three measures separately.
2. POPULARITY INFERENCES were elicited using three measures: ratings on a 7-point scale,
as well as the two open-ended measures, space allocation and relative market share
(both to sum to 100 across the five options), combined to form the popularity index.
3. ATTENTION was measured in the following manner:
a. Participants were given the following instructions:
“Packaging is an important component of the product design process, and managers are
constantly attempting to improve their packaging. Circle the package that you think is
the most likely to be the first one noticed. This is the packaging that you think stands
out the most, to which your eye was first drawn when you looked at the category, and
the one you noticed the most.”
b. Subsequently, to maintain the cover story, participants were asked to rate the overall
attractiveness of each of the packages (1=Not at all attractive/ 7 = Very attractive),
and then asked how attention-getting each of the package designs were (1=Not at all/
7 = Very attention-getting). An exploratory factor analysis across these measures
revealed a three factor solution (68.64% variance explained), such that the measures
of attractiveness and attention for the 1st and 3rd position loaded onto the first factor,
the measures of attractiveness and attention for the 2nd and 5th position loaded onto
the second factor, and the two measures for the 4th position loaded onto the third
factor. Given this factor structure the two variables have been reported separately.
22
c. Finally, we asked respondents to rate the extent to which they agreed with
statements regarding how much attention they pay to products placed in different
positions. These statements were: “I always look at the first item in a display,” “I do
not pay much attention to the last choice given,” and “I always look at the middle
option first” (1= disagree/ 5 = agree).
Results. Results, by measure, are presented in Table 1 and described below.
-- Insert Table 1 about here. –
RANK. As the order is the same across all conditions, rank was treated as a continuous
variable (with a lower number indicating a higher rank and higher preference), and subjected to a
five level repeated measures ANOVA incorporating order as a between-subjects variable. As
predicted, study participants ranked the option in the middle highest (M = 2.57 where 1 = highest
rank and 5 = lowest rank), with lower preferences for options in the extreme positions (extreme
left = 3.26, extreme right = 3.45), as compared to those closer to the center (left of center = 3.04,
right of center = 2.68). This was reflected in a main effect of position (F (4, 520) = 8.00, p < .001,
η
2 = .06), with a significant quadratic contrast reflecting the U-shaped curve of the mean ranks
(F(1, 130) = 41.25, p < .001,
η
2 = .24).
PURCHASE INTENTIONS. A 5-level repeated measures ANOVA on purchase intentions
also revealed a main effect of position (F (4, 520) = 8.33, p < .001,
η
2 = .06) with a significant
quadratic contrast (F (1,130) = 30.24, p < .001,
η
2 = .19) The means follow an inverse U-shaped
curve with higher numbers reflecting higher intentions (Ms = 3.36, 3.77, 4.34, 4.43, and 3.83 for
the 1st - 5th position respectively). There was a main effect of order (F (5, 130) = 2.27, p = .051,
η
2 = .08) reflecting highest intentions in the two order conditions when choice was the first
measure elicited (Ms = 4.22 and 4.38), followed by when it was the second measure elicited
23
after popularity ratings (M = 4.03), or after attention ratings (M = 3.85), and lowest ratings when it
was elicited last (Ms = 3.60 and 3.63). Order did not moderate the position effect.2
LIKING. A repeated measures ANOVA on liking also revealed a main effect of position (F
(4, 516) = 6.06, p < .001,
η
2 = .04) with a significant quadratic contrast (F (1,129) = 19.02, p < .
001,
η
2 = .13) The means follow an inverse U-shaped curve with higher numbers reflecting
greater liking (Ms = 3.51, 4.02, 4.19, 4.55, and 3.88 for the 1st - 5th position respectively).3
POPULARITY RATINGS. An ANOVA on the popularity index revealed a position effect (F
(4,520) = 9.83, p < .001,
η
2 = .07), with a significant quadratic contrast (F (1,130) = 52.39, p < .
001,
η
2 = .29). The means are in the direction predicted by H2: lowest popularity ratings for
varieties at the extremes (Left = 36.52. Right = 42.60), followed by varieties in the middle
positions (Ms = 46.14, 48.32 and 49.43 for left of center, center, and right of center respectively).
MEDIATION OF PURCHASE INTENTIONS BY POPULARITY. The repeated measure ANOVA on
purchase intentions incorporating the popularity indices as covariates showed that the effect of
position reduced to non-significance (F = .21, p = .93,
η
2 = .002), with all covariates significant
(F(1, 124) = 11.56, 13.19, 13.06, 16.06, and 15.17, p < .001 for all,
η
2 = .09, .10, .10, .11, and .
11 for indices 1-5 respectively). This reflects a pattern of perfect mediation (Baron and Kenny
1986). The main effect of order remained significant (F (5,124) = 2.29, p < .05,
η
2 = .08).
2 To confirm that the measure of purchase intentions mediated ranks assigned to the different items, we reran the
repeated measure ANOVA on the five ranks, incorporating the five measures of purchase intentions as covariates.
This analysis showed that the previously significant effect of position reduced (F (4, 496) = 5.11, p < .001,
η
2 = .04),
with all covariates significant (F(4, 496) =24.22, 34.62, 32.21, 59.40, and 25.96, p < .001 for all,
η
2 = .16, .22, .21, .32,
and .17 for intentions 1-5 respectively). This is consistent with partial mediation (Baron and Kenny 1986), though the
order of the two measures was not counterbalanced. Given this pattern, however, later tests for mediation to test H3
are only conducted with the purchase intention variable (vs. both the rank and the purchase intention variables).
3 To confirm that the measure of liking mediated purchase intentions, we reran the repeated measure ANOVA on
purchase intentions, incorporating the five measures of liking as covariates. This analysis showed that the previously
significant effect of position was no longer significant (F = .24, p = .91,
η
2 = .002), with all five covariates of liking
involved in significant interactions with the position effect (F(4, 492) = 50.04, 38.09, 40.56, 89.55, and 39.37, p < .001
for all,
η
2s = .29, .24, .25, .42, and .24 for liking 1-5 respectively), as well as exerting main effects (F(1, 123) =8.77,
8.06, 8.02, 22.49, and 9.08, p < .005 for all,
η
2s = .07, .06, .06, .15, and .07 for liking 1-5 respectively). The main effect
of order continued to be significant (F(5, 123) = 4.03, p < .005,
η
2 = .14). This implies that intentions are based on
respondents’ own liking for the products and feed into their choice behavior: a result we return to in studies 5 and 6.
24
ATTENTION. A majority of respondents identified the product in the fourth position as the
most noticeable (65/ 82), with the lowest percentage identifying the product in the middle
position as the most noticeable (1/82; 1st = 8/ 82 2nd = 4/ 82, and 5th = 4/ 82, non-response = 54,
χ
24 = 181.54, p < .001). Responses to the statements regarding where attention was directed
revealed greater agreement with the statement that attention is directed to the first item in a
display (M = 3.04), than with either the statement that attention is directed to the center of the
display (M = 2.57, paired t = 5.02, p < .001), or the statement that people do not pay attention to
the last choice given (M = 2.40, paired t = 3.45, p < .001). The level of agreement with the latter
two statements did not differ (paired t = 1.37, p = .174). Thus, direct reports of attention do not
corroborate the explanation that middle positions are attended to more than others.
Further, a five-level repeated measures ANOVA on the direct rating of how attention-
getting each of the packages was revealed a significant main effect of position (F (4,524) =
74.98, p < .001,
η
2 = .36), with linear, quadratic, cubic and order 4 contrasts all significant (F
(1,131) = 11.26, 35.64, 150.50, and 111.34, ps < .001 for all,
η
2s= .08, .21, .53, and .46). The
pattern of the means reveals that the package in the position right of center was rated as most
attention-getting (M= 5.73), followed by the 1st three positions (Ms = 3.80, 3.85, and 3.75), with
the extreme right position rated as least attention-getting (M= 3.39).4
4 The pattern is similar for the ratings of the attractiveness of the five packages (Ms = 3.95, 3.96, 4.01, 5.54, and 3.69
for positions 1-5, from left to right respectively, F(4, 524) = 46.32, p < .001,
η
2 = .26) with 1st-4th order contrasts
significant (F (1,131) = 11.07, 33.02, 72.60, and 53.06, ps < .001 for all,
η
2s= .08, .20, .36, and .29 respectively). This
analysis also revealed an interaction of order with attractiveness ratings (F (20,524) = 1.85, p < .05,
η
2 = .07),
reflecting a smaller effect of position (specifically, difference between the rating of the attractiveness of the fourth
position versus the remaining positions) when these ratings were the first to be elicited. Even though the order of the
ratings of attention and attractiveness was not counter-balanced, to examine if attractiveness ratings mediated the
effect of position on attention, we reran the repeated measure ANOVA on the five reports of attention, incorporating
the ratings of attractiveness as covariates. This analysis showed that the effect of position remained significant,
although lower in size (F(4, 504) = 3.32, p < .05,
η
2 = .03), with the main effect of attractiveness ratings for all
positions except the middle position significant F(1, 126) = 14.88, 10.81, 14.13, and 7.21, p < .01 for all,
η
2s = .11, .08,
.10, and .05 for attractiveness ratings 1,2, 4 and 5), and all ratings of attractiveness involved in a significant interaction
with the position effect (F(4, 504) = 19.16, 23.11, 9.18, 14.92, and 14.52, p < .001 for all,
η
2 = .13, .15, .07, .11, and .
10 for attractiveness ratings 1-5 respectively). This reflects a pattern of partial mediation (Baron and Kenny 1986)
tentatively suggesting that attention may be driven by attractiveness of the package design rather than by the position
on the package in an array. However, this result would have to be replicated after counterbalancing the order of the
measures and including filler tasks in between them prior to any definitive conclusions being drawn.
25
To examine if attention mediated the effect of position on purchase intentions, we reran
the repeated measure ANOVA on the five purchase intentions, incorporating the ratings of
attention as covariates. This analysis showed that the effect of position reduced to non-
significance (F = .47, p = .76,
η
2 = .004). The interactions between the amount of attention paid
to the two left most items and the position factor were the only other significant effects (F(4, 500)
= 5.10, and 2.38, p < .05 for both,
η
2s = .04, and .02 for extreme and left of center respectively).
As these results could be consistent with mediation, we examined whether incorporating both
ratings of attention and the popularity indices in the repeated measures ANOVA would identify
which of these constructs mediated the route to higher purchase intentions. This analysis
showed that the main effect of position was again not significant as in previous ANCOVAs (F = .
09, p = .99,
η
2 = .001) but neither were the main and interaction effects involving the attention
ratings. However, the main effects of the five popularity indices were significant (F(1, 119) =
7.52, 9.91, 10.38, 11.93, and 10.09, p < .01 for all,
η
2s = .06, .08, .08, .09, and .08 for popularity
indices 1-5 respectively). This reflects a pattern of perfect mediation of intention via popularity
inferences ruling out attention as an alternative mediating mechanism (Baron & Kenny, 1986).
DISCUSSION. To summarize, Study 3 replicated the results of studies 1 and 2 using a richer
set of dependent variables to demonstrate the preference for the central position. Additionally,
study results replicated study 2 effects ruling out attention as a mediating mechanism behind the
centre-stage effect. Finally, it tested H2 and H3 to show that the preference for the central
position is mediated by beliefs regarding the popularity of the central item.
Studies 1-3 demonstrate that consumers use the physical placement of a product as a cue
to infer its popularity and base their choices on this inference. This is reflected in the fact that
their intentions are based on popularity ratings. However, a belief-based inference mechanism
also implies that position effects should be contingent on the use of alternate sources of
information to make a judgment. Specifically:
26
I. It implies that discounting the diagnosticity of position should attenuate the effect. Study
4 tests this hypothesis (H4).
II. It also implies that even when position is used as a cue to infer popularity, the effect of
popularity on purchase intentions should be contingent on the relative importance of
satisfying the preferences of others (for which popularity would be a diagnostic cue)
versus satisfying one’s own preference. Study 5 tests this effect (H5).
III. Finally, it implies that position as a cue to infer popularity may not be used as much if
alternative individuating information becomes accessible. Study 6 tests whether
increasing the accessibility of one’s own attitudes prior to choice reduces the effect of
position as a cue to make product judgments (H6).
These three studies, investigating the conditions that attenuate the use of belief-based
processing, to identify boundary conditions of the center-stage effect, are now described.
Study 4: Discounting the Diagnosticity of Position as a Cue
The goal of this study was to examine whether the center-stage effect can be attenuated
when position is no longer informative (H4). As argued earlier, the strength of consumer
inferences based on physical position will depend on both the perceived diagnosticity of shelf
positions as information as well as the accessibility of alternative individuating information.
Specifically, a shelf space position should be perceived to be more diagnostic if it reflects a
chosen retailer strategy. On the other hand, if a shelf space array was arranged at random, there
is less reason to believe that shelf space positions are informative of brand popularity. If belief
processing regarding shelf space arrays, drives the centre stage effect, then manipulating the
information value of a shelf space array should moderate the centre-stage effect: it should be
stronger when arrays are believed to reflect retailer policy as compared to a random order. On
the other hand, if the centre-stage effect is due to attention paid to the items in the centre, then
manipulating the information value of the arrangement of a shelf-space should not affect choice
27
behavior. Study 4 investigates this issue to garner further support for the hypothesis that centre-
stage effects are due to belief processing.
Method. The scenario used was identical to that in Studies 2 and 3 with the following
differences: i) In order to make the choice scenario better reflect the length of arrays that
consumers may encounter, nine (rather than five) varieties of pretzels were shown using nine
different presentation orders; ii) A between-subjects manipulation explained that either “the
presentation order had been selected at random” or that it “represented real product placement
by a local retailer.” The dependent variable was the choice of the variety in the middle position.
Results. The choice of the middle option was contingent on whether participants were told
that the ordering of options was random versus real (χ28 = 15.03, p < .05). While 4.5% chose the
middle option in the “random” condition, as many as 31.8% did so in the “retailer” condition (p < .
05 versus chance odds of 1/9 or 11.1%). Thus, explicitly discrediting the diagnosticity of position
as a cue to make judgments reduces the reliance on position as a source of information,
supporting belief processing as the underlying reason for centre-stage effects.
Study 5: Manipulating the Diagnosticity of Popularity as an antecedent of intention
The goal of this study was to test H5. We examine whether the center-stage effect is
greater when people are making purchases for others with unknown preferences and asked to
consider the specific reasons why other people would or would not like each of the products in
their choice set. These conditions should increase the relevance of using popularity inferences to
make purchase decisions, as compared to when respondents are making purchases for
themselves and are asked to consider the specific reasons why they like or dislike each of the
products in their choice set. In Study 3, we noted that intentions were based on a respondents’
liking for the products that fed into their choice behavior. In this study we used a moderation
approach to test whether intentions to purchase the product in the centre are higher when
respondents’ think about others’ as compared to their own preferences.
28
A second goal of this study was to replicate the mediating paths noted in Study 3, using
not just one order of presentation of the stimuli, but five different orders as in Studies 2 and 4.
Third, we examine a more powerful set of measures to rule out the alternative explanation that
attention (rather than popularity inferences) leads to the centre-stage effect.
Method. We used the same scenario as Studies 2-4, using the complete set of measures
from Study 3, and counterbalanced the order of the five products. Unrelated tasks were
administered in between the different sets of dependent measures as filler tasks to erase short
term memory and reduce consistency pressures and order effects. The measures were elicited
in the following order: choice, attention, popularity, and recall.
We introduced a between subjects manipulation of the amount of consideration given to
others’ preferences instead of one’s own preferences. In the “others’ preferences” condition,
respondents were told “You are shopping for a party where guests need to bring a snack that
everyone will like”. In the “own preference” condition, participants were told “You are shopping
for a party where guests need to bring a snack that is their personal favorite”. Before making
their choices, respondents were asked to “Think about why or why not each of the varieties will
be liked by other people” (other’s preferences) or “Think about why you like or dislike each of the
flavors above” (own preferences).
Participants rank ordered their preference for each variety, wrote down the reasons why
they made each choice, and completed the intention and liking questions for all five varieties
(using the same 7-point scales). Exploratory factor analyses showed that liking and intentions for
each product loaded on to the same factor in both conditions (Variance explained = 87.82% and
92.87% for other and self respectively; factor loadings > .89 for each item on each factor). Given
this structure (and the similar results for the two measures in Study 3), the two measures were
combined into an “evaluation index” for each of the five products (r between intent and liking > .
72 for all five products separately in both conditions).
29
After completing an unrelated task, participants answered the attention questionnaire that
measured which package was the most noticeable, ratings of package attractiveness and
attention-getting, as well as their level of agreement with the three statements regarding whether
or not they paid attention to the first, middle and last option in an array. After another filler task,
respondents completed the popularity-inferences questionnaire with the three measures
comprising the popularity index: popularity ratings, space allocation and relative market share.
Finally, after completing a third filler task, respondents were administered a surprise recall
task as in Study 2. The free recall was coded in terms of whether or not the product in a given
position was recalled, with a subset also coded for the order in which the different positions were
recalled (e.g., if the first item recalled was in the middle position, it was coded as a 3, but if it was
in the first position it was coded as a 1). We used the 5-point VVIQ response scale to ask
participants to rate how clearly they were able to visualize each of the different packages given
only the name of the package.5
Results. Results, by measure, are presented in Table 2 and described below.
-- Insert Table 2 about here. –
RANK. A cross-tabulation of the first two choices by position (centre/ all others) by condition
(self/ other) revealed a significant interaction (
χ
2 = 4.77, p < .05). Whereas 40/74 (or 54.54%,
binomial p < .01 versus chance percentage of 40%) of respondents in the condition where they
were making a choice for others chose the product placed in the middle position, only 36.36%
(or 28/77, n.s. versus 40% chance likelihood) did so in the condition where they had to choose
their personal favorite. Thus, asking respondents to consider their personal preferences
successfully attenuated the centre-stage effect; while the effect was replicated in the condition
where they considered others’ preferences.
5 The response categories are: 1= Perfectly clear and as vivid as normal vision; 2= Clear and reasonably vivid; 3=
Moderately clear and vivid; 4= Vague and dim; 5= No image at all, you only "know" that you are thinking of the object.
30
PURCHASE INTENTIONS AND LIKING INDEX. A 5-level repeated measures ANOVA using
self/ other as a between subjects factor revealed a significant effect of condition (F (1,149) =
6.08, p < .05,
η
2 = .04), with no other effects significant. Given the main effect of condition and
the directional hypothesis regarding the centre-stage effect, we examined whether evaluations of
the product in each position differed in the two between subjects conditions. T-tests of
differences in means across conditions for each of the five positions, revealed that the only
position where evaluations differed was the central position (M = 8.97 vs. 7.68 for other versus
self respectively, t149 = 2.16, p < .05). In the remaining four conditions mean evaluations were not
significantly different from each other (t149 = 1.07, 1.80, .21, and 1.12 for the two left and the two
right positions respectively, p > .05 for all).
POPULARITY INDEX. A repeated measures ANOVA on the popularity index incorporating
self/ other as a between subjects variable revealed a quadratic contrast for position (F (1,148) =
4.66, p < .05,
η
2 = .03), reflecting an inverse-U shaped curve (Ms = 43.59, 44.54, 46.92, 46.06,
and 42.18 from extreme left to extreme right respectively). The condition factor did not exert a
main or interaction effect, suggesting that inferences of popularity were made in both conditions.
ATTENTION. There were no differences across positions in the identification of the product
that was most noticeable (26, 32, 30, 28 and 23 people out of 139, non-response 12, identified
the product in the 1st-5th position from left to right as the most noticeable,
χ
24 = 1.76, p > .75). A
five-level repeated measures ANOVA on the direct rating of how attention-getting each of the
packages was, and a similar analysis on the ratings of attractiveness, including self/ other as a
between-subjects condition, also revealed no significant effects. These results suggest that the
preference for the fourth position in Study 4 can be attributed to the use of a single order of
presentation, rather than being a robust effect.
Responses to the statements regarding where attention was directed revealed greater
agreement with the statement that attention is directed to the first item in a display (M =2.84),
31
than with either the statement that attention is directed to the center of the display (M = 2.61,
paired t = 1.75, p < .09), or the statement that people do not pay attention to the last choice
given (M = 2.52, paired t = 3.13, p < .01). The level of agreement with the latter two statements
did not differ (paired t = .81, p = .42). These results replicate those of Study 3 and show that
people do not believe that they pay more attention to products in middle positions.
As in study 2, there was no difference in the likelihood of recalling a product across
positions (n = 122, 116, 123, 127, and 121 out of 151), which was high for all positions. A subset
(n = 94, or 62.3%) of the recall responses were coded in terms of which position was the first
recalled (second recalled, etc.) to examine top-of-mind recall as a surrogate for attention. The
first item recalled was most likely to be the products in the first position (n = 33, 16, 12, 20, and 9
for positions 1-5, no item recalled = 4,
χ
24 = 19.44, p < .001). There were no differences across
position in the second item recalled (n = 18, 24, 18, 10, and 17 for positions 1-5, no item recalled
= 7,
χ
24 = 5.70, p = .22); with the central position being more likely to be recalled as the third item
(n = 7, 14, 27, 19, and 16 for positions 1-5, no recall = 11,
χ
24 = 12.84, p < .05). There were no
differences across position for the 4th item (n = 17, 11, 11, 16, and 21 for positions 1-5, no recall
= 18,
χ
24 = 4.79, p = .31) or the 5th item recalled (n = 8, 12, 11, 16, and 15 for positions 1-5, no
recall = 32,
χ
24 = 3.32, p = .50). Across these measures it appears that the order in which items
were recalled reflected also the order in which they were presented.
Therefore, to further examine whether there was any difference in the clarity with which
items in different positions were recalled, we examined the reported vividness of visual images
for the five positions including self/ other as the between subjects condition in a repeated
measures ANOVA. This analysis did not reveal any significant effects across position (Ms =
2.82, 3.05, 3.05, 3.08, and 3.05 for positions 1-5 from left to right respectively, with lower
numbers indicating more vivid visual images). Therefore, using this measure as well, there is no
support for the idea that items in the centre attract greater attention.
32
However, it remains possible that the position that was more noticeable was the position
more likely to be chosen. To examine this explanation, we cross-tabulated the top-ranked choice
(middle option/ other options) with whether or not the middle option was identified as being most
noticeable (yes/ no) for the two between-subjects conditions separately. When respondents were
asked to make a choice that others would like, there was no relationship between whether or not
the middle option was chosen and which position had been identified as the most noticeable (
χ
2
= .99, p = .32). On the other hand, when respondents were asked to consider their own
preferences, then 6/15 respondents chose the middle option when they identified it as being
most noticeable, as compared to 8/53 choosing it when they identified other options as being
most noticeable (
χ
2 = 4.44, p < .05). (The same analysis conducted using all five positions rather
than combining positions into centre/ not centre revealed an identical pattern: No relationship
between notice-ability and top ranked choice in the “others” condition:
χ
216 = 16.35, p = .43, and
a significant relationship in the “self” condition:
χ
216 = 40.26, p < .001). Therefore, when
respondents are selecting for themselves, the attention they pay to the products is related to
their choice, but when they are making choices for other people, then, it is not. Irrespectively, the
fact that other options were identified as being more noticeable than the middle option, rules out
that attention drives the preference for the centre-option. Further evidence for this path is
examined using a moderated mediation approach described below.
MODERATED MEDIATION ANALYSES. To examine the route to evaluations, we first
examined whether those who chose the central positions as one of their first two top-choices had
a different pattern of evaluations (purchase intentions and liking index) across the positions, for
both the “others” and “self” condition separately. A 5 (positions) x 2 (choice of centre product or
not) repeated measures ANOVA revealed significant position x choice interactions in both
conditions (F(4, 288) = 13.43, and F(4, 300) = 12.65 for others and self respectively, p < .001 for
both,
η
2 = .16 and .14). The interaction reflected a higher evaluation for the middle position when
33
it was the one chosen in both conditions (Ms = 11.20 vs. 6.35 for others, and 10.89 vs. 5.84 for
self, t = 7.49 and 7.60, p < .001 for both).6 Thus, evaluations of the middle option are higher
when it is chosen.
A series of ANCOVAs explore the route to these higher evaluations using only popularity
indices as covariates, only attention ratings as covariates, and then, including both popularity
indices and attention ratings together, to test whether the route to improved evaluations is driven
by popularity in the others condition (controlling for attention), but not in the “self” condition:
The ANCOVA including popularity indices as a covariate showed that in the “others”
condition the choice x position reduced in effect size (F(4, 268) = 5.87, p < .001,
η
2 = .08 [vs.
η
2
= .16 without the covariate]), with main effects for all five popularity indices (F(1, 67) = 10.19,
8.33, 10.90, 11.93, and 10.48, p < .01 for all,
η
2s = .13, .11, .14, .15, and .13 for 1-5 respectively)
and a position x popularity index2 interaction (F(4, 268) = 3.77, p < .01,
η
2 = .05). In the “self”
condition the choice x position effect size was also reduced (F(4, 276) = 9.88, p < .001,
η
2 = .12
[vs.
η
2 = .14 without the covariate]), and there were significant position factor interactions with
the popularity indices for positions 2 and 5 (F(4, 276) = 3.77 and 3.45 respectively, p < .01 for
both,
η
2s = .05 for both).
The same analyses incorporating attention ratings as a covariate showed that in both the
“others” condition (F(4, 268) = 11.03, p < .001,
η
2 = .14), and the “self” condition, the choice x
position effect remained significant (F(4, 276) = 11.36, p < .001,
η
2 = .14), with no significant
main or interaction effects involving the covariates of attention-ratings.
Finally, incorporating both the popularity indices as well as ratings of attention in the
evaluations ANOVA showed that:
6 There was no difference for the 4th (Ms = 7.65 vs. 8.29 for others, and 7.89 vs. 7.78 for self, t = .77 and .13, p > .44
for both) or 5th position (Ms = 8.35 vs. 8.65, t = .37 for others, and 7.19 vs. 8.20, t = 1.16 for self, p > .25 for both).
Evaluations were lower for the first position in the “other” condition when the middle option was chosen (Ms = 7.10 vs.
9.29, t = 2.74, p < .01), and no different in the “self” condition (Ms = 6.46 vs. 8.02, t = 1.72, p = .09), with the reverse
pattern true for the evaluation of the option in the second position (Ms = 7.90 vs. 8.68, t = .95, p > .34 for others, and
6.04 vs. 7.88 for self, t = 2.17, p < .05).
34
i) In the “others” condition the choice x position effect was significant (F(4, 248) = 5.08, p
< .001,
η
2 = .08), with main effects for all five popularity indices (F(1, 62) = 5.78, 4.46, 7.97, 6.26
and 6.15, p < .05 for all,
η
2s = ..08, .07, .11, .09 and .09 for popularity indices 1-5 respectively)
and a continued position x popularity index2 interaction (F(4, 248) = 3.40, p < .01,
η
2 = .05).
Thus, popularity inferences, rather than attention, mediated the route to evaluations in the
“others” condition as predicted.
ii) In the “self” condition, the choice x position effect remained significant (F(4, 252) =
9.35, p < .001,
η
2 = .13). There were no main effects of any covariates, but the position factor
was involved in significant interactions with the popularity indices for positions 2 and 5, as well
as attention paid to the middle item (F (4, 252) = 3.42, 3.32, and 2.45 respectively, p < .05 for all,
η
2s = .05, .05 and .04). Thus, both popularity inferences as well as attention to the central item
affect preferences for the middle item when people consider their own preferences.
DISCUSSION. To summarize, the moderated mediation analyses show that the path to
improved evaluations for the middle position when people are buying for others is via popularity
judgments rather than attention paid to the different positions; whereas, when they are buying for
themselves, popularity inferences are weaker antecedents of evaluations that are also based on
attention paid to the middle position. In this study the overall position effect on evaluations was
weaker than in previous studies. This could be because the procedure asked participants to
think about the reasons for their choice. The greater amount of thought paid to the choice task
could have led to an attenuated use of the position cue for judgments. The next study examines
this idea.
Study 6: Manipulating the Accessibility of Individuating Information
Contextual information is less likely to be used when individuating information is
accessible, and, therefore, a primary input into a choice decision (Fiske & Neuberg, 1990). Study
35
6 tests this hypothesis by contextually manipulating the accessibility of a prior evaluation by
changing the order in which choices and product judgments are elicited. When judgments are
made prior to choice then the accessibility of consumer product evaluations is greater during the
choice task, reducing the effect of information inferred from the position of products in the choice
set (Feldman & Lynch, 1988).
Additionally, choice is inherently comparison based whereas evaluation leads decision
makers to consider the absolute values of product options (Nowlis & Simonson, 1997). As a
consequence, choosing first should produce a higher reliance on position and spatial cues. This
prediction (H6) is tested below.
Method. We used the same scenario as in Study 1: offering participants a packet of
chewing gum as a token of appreciation for their taking part in an unrelated study. There were
three packets of Extra Chewing gum used: Spearmint, Peppermint, and Winterfrost, presented
using three different orders (SPW, WSP and PWS). However, unlike Study 1, participants were
given a questionnaire where the picture of the three chewing gum packets was reproduced. All
participants indicated which chewing gum they would choose. On a separate page they rated the
attractiveness of the package of each of the chewing gums (1 = Do not like at all/ 7 = Like a
whole lot), the quality of each of the flavors (1 = Poor quality/ 7 = Good quality), and how popular
each of the flavors were, that is, the extent to which the average consumer will like each of the
flavors (1 = Not at all popular/ 7 = Very popular). An exploratory factor analysis on the nine
ratings revealed a three factor solution, with each factor including the three ratings for the
product (variance explained = 64.26%, factor loadings > .74 for all). Therefore, the three ratings
were combined into an evaluation index separately for each position (α = .73, .74 and .67,
respectively). The between-subjects manipulation was the order in which choice was elicited: In
one condition choice preceded ratings, whereas in the other condition it followed it.
Subsequently, participants rated their level of agreement using a 5-point scale with three
position-related belief statements referring to product choice: “Most popular products are always
36
in the middle,” “I always choose the first item in a display,” and “I do not pay much attention to
the last choice given” (1 = Disagree/ 5 = Agree).
Results. Results, by measure, are presented in Table 3 and described below.
-- Insert Table 3 about here. –
CHOICE. A cross-tabulation of the number of times the chewing gum in the centre was
chosen as a function of whether attitudes were formed before or after the choice showed a
significant interaction (
χ
2 = 6.55, p < .01). When choice preceded attitudes, the product in the
centre was more likely to be chosen (45.2%), replicating Study 1 results. However, the centre-
stage effect was attenuated when consumers had rated the chewing gums prior to making their
choices: 26.7% chose the gum in the middle; a share that is no different from chance likelihood.
EVALUATION INDEX. A repeated measures ANOVA on the three evaluation indices using
order of elicitation (choice first/ evaluations first) and choice (1st, middle or last option) as
between-subjects factors revealed a significant order of elicitation by choice interaction (F (4,
334) = 33.89, p < .001,
η
2 = .29), implying that choices and evaluations were aligned.
POSITION RELATED JUDGEMENTS. Participants had the highest level of agreement with the
belief “Most popular products are always in the middle” (M = 2.73, on a 5-point disagree-agree
scale), followed by agreement with the statement “I do not pay much attention to the last choice
given” (M = 2.19, paired t = 4.60, p < .001), which in turn was higher than the level of agreement
with the statement “I always choose the first item in a display” (M = 1.91, paired t = 3.29, p < .
001). Order of choice/ ratings did not affect judgments (F < 1 for all).
MEDIATION ANALYSIS. As beliefs were not affected by order of elicitation, we can examine
the strength of the relationship between choices, evaluations and beliefs in the two order
conditions. We conducted a logistic regression on the dependent variable of whether or not the
middle option was chosen, incorporating the three evaluation indices and the three position-
related beliefs as explanatory variables in the analysis, separately for the two order conditions. In
37
the condition where choice was elicited first, the effect of the three evaluation indices was
significant (Wald = 5.22, 14.23, and 9.57, p < .05 for all), as was the effect of the belief that
popular choices are in the middle of a display (Wald = 4.82, p < .05). On the other hand, when
choices were elicited after evaluations, then only the effect of the first two evaluation indices was
significant (Wald = 6.57 and 13.19, p < .05 for both), whereas the effect of the belief that popular
choices are in the middle of a display was not significant at p > .05.
Discussion. To summarize, the results of this study provide further evidence that the
centre-stage effect is driven by a belief substitution of internally formed individuating information
about the brand for contextually constructed one based on the position of the brand in an array.
In particular, consumers seem to rely on position-based beliefs to a greater extent when they
make their choices prior to considering (via making evaluations) their preferences for the
product. These findings both clarify Study 5’s results and support H6.
GENERAL DISCUSSION
This paper investigated the reason for a central position advantage, or “centre-stage”
effect, in product choice. Findings support the contention that the “centre-stage” advantage is
due to inferences about the higher popularity of products placed in the middle of an array.
Results from six experiments show that i) there is a centre-stage effect: a product in a central
position is preferred over one at either end of the array; ii) the centre advantage is due to beliefs
that the product in the centre is the most popular one rather than due to greater attention paid to
the product, iii) the centre advantage is attenuated when position does not convey information,
the information is less relevant for choices, and when people’s intrinsic product preferences are
made accessible by the context.
Theoretical Implications
38
Our proposed belief-based model parsimoniously explains earlier findings by Christenfeld
(1995) and Shaw (2000), which showed consumers’ preference for the middle option among an
array of identical items. Our results replicate Shaw’s earlier findings that attention does not drive
the effect. Instead, we find that position is used as a cue to infer the popularity of a brand, which
follows through to product preferences. This belief-based model also allows for an explanation of
different order effects documented in seemingly unrelated streams of research dealing with
placement, whether the placement is of people, response alternatives, choices or products. In
particular, these results speak to the literature in the following domains:
Packaging and Placement Beliefs. In a package design task, consumers may be more
used to seeing brand name information on the right and logos on the left, and, therefore, when
the design is consistent with their prior expectations, the package receives more attention and is
better recalled (Janiszewski, 1990). In a store display, consumers expect that products located in
an end-of-aisle display are also on discount leading to a sales spike for these products
regardless of whether they are actually discounted or not (Inman et al., 1990).
People Placement Beliefs. The spatial belief heuristic is consistent with McArthur &
Post’s (1977) speculation that prominent positions are evaluated more favorably due to learned
cultural norms (Raghubir & Valenzuela, 2006, Taylor & Fiske, 1975).
Survey Method Beliefs. A potential reason for the order effects found in impression
formation (Asch, 1946), and survey methods (Schuman & Presser, 1996) is that individuals and
respondents may simply follow conversation norms (Grice, 1975), assuming that the reason an
item was provided first (or last) was because it was most diagnostic of a personality (in an
impression formation task), or because it is most likely to be chosen by others (in a survey
methods context). That is, learned norms regarding placement order dictate people’s responses,
and these norms are situation and context specific. This is consistent with our proposed spatial
belief theory where the order of placement of an item in an array is in itself informative, with the
content of the information represented by the position contingent on specific learned norms.
39
Managerial Implications
The internet revolution has allowed for online shopping to be a well-accepted way of
purchasing products. Previous research has examined the influence of unique online store
characteristics, such as interactive decision aids and customization procedures (Senecal &
Nantel, 2004; Simonson 2005). Less attention has been paid to the equally intriguing question of
whether the same traditional marketing principles affect online purchase decisions as well. In a
busy retail environment with millions of brand choices and limited information processing ability,
the effect of position cues is particularly relevant. Our research can provide some insights about
the effects of a product’s physical placement on a Web display. Online purchase decisions seem
to be an appropriate context in which to apply our findings since space allocation is rarely an
issue and limited eye movements are usually enough to scan the entire display (Breugelmans,
Campo & Gijsbrechts, 2007). Our results support that shelf organization does have an effect on
consumers’ purchase decisions. Since we find that a position advantage is due to improved
attitudes towards the brand, that these attitudes are not mediated by attention, but rather reflect
an overall belief, this implies that managers should be willing to pay a premium to have their
brands occupying the premium position in an online product layout. Web retailers should be able
to use this information to bargain with manufacturers for shelf benefits or promote brands that
are not market leaders. However, our findings also show that the “centre-stage” effect can be
attenuated by making pre-existing attitudes or other diagnostic information accessible (e.g.,
using a “Market Leader” label on packaging).
Study Limitations and Areas for Future Research
Even though consumers seem to apply general rules in terms of the value of different
positions in a retail context; shopping environments differ in the extent of distraction and
stimulation they provide (e.g., aural stimuli, flashing lights, or other shoppers). As a
40
consequence, a relevant question is whether such competing demands on a consumers’ time
would ameliorate or exacerbate the centre-stage effect. Theoretically, this can be examined by
seeing if the effect increases when consumers are under cognitive load. If the effect increases
under load, this suggests that the use of the position belief does not consume cognitive
resources (i.e., may be automatic), but if it were to decrease, it would imply that the centre-stage
belief requires conscious cognitive resources to be used.
Future research could also examine the moderating role of individual difference to assess
the robustness of the effects to differences in the depth of processing as measured by people’s
need for cognition as done in the Inman et al. (1990) study, the use of visual information as a
cue to make judgments as measured by the vividness of visual imagery (Marks 1972), and the
importance of satisfying others as compared to oneself as measured by the cultural difference
individualism-collectivism scale (e.g., Markus & Kitayama, 1991).
Additionally, not all choices among similar options have been characterized by centrality.
For example, Christenfeld (1995) found a preference for extremity when people chose a route
among several equivalent ones. Besides, studies on order effects in sequentially presented
information have typically indicated that the first or last items in a sequence are more influential
than the middle ones (Nisbett & Wilson, 1977). It is possible an extremity advantage exists for
choices containing sequentially presented options, whereas a centrality advantage exists for
simultaneously presented ones. That is, with sequentially presented options there may be a
different conversation norm or market-place meta-cognition. Specifically, when respondents
need to choose from a large range of options, then it is plausible that they expect that the most
popular options will be the first or the last ones to be presented. This belief has been proposed
as an explanation for recency and frequency effects in survey methods (Asch, 1946; Schuman &
Presser, 1996). It is possible that the sequential presentation of nominal alternatives in the
Nisbett and Wilson (1977) study triggered this belief as applied to marketplace meta-cognitions
in a choice context, leading to the recency effect where consumers chose the last option
41
presented in a series of stockings. Research should further test whether the sequential or
simultaneous presentation mode moderates whether choices reflect a centre-stage effect versus
recency or primacy effects.
42
References
Asch, S. E. (1946). Forming impressions of personality. Journal of Abnormal and Social
Psychology, 41, 258-290.
Attali, Y., & Bar- Hillel, M. (2003). Guess where: The position of correct answers in multiple-
choice test items as a psychometric variable. Journal of Educational Measurement, 40(2),
109-128.
Baron, R. M., & Kenny, D.A. (1986). The moderator-mediator distinction in social psychological
research: Conceptual, strategic, and statistical considerations. Journal of Personality and
Social Psychology, 51(6), 1173-1182.
Breugelmans, E., Campo, K., & Gijsbrechts, E. (2007). Shelf sequence and proximity effects on
online grocery choices. Marketing Letters, 18(1-2), 117-133.
Buchaman, L., Simmons, C. J., & Bickart B. A. (1999). Brand equity dilution: Retailer display and
context brand effects. Journal of Marketing Research, 36, 345-355.
Chaiken, S. (1987). The heuristic model of persuasion. In M. Zanna, J. Olson, & C. Herman
(Eds.), Social Influence: The Ontario Symposium, vol. 5 (pp. 3-39). Hillsdale, NJ: Lawrence
Erlbaum.
_________, & Maheswaran, D. (1994). Heuristic processing can bias systematic processing:
Effects of source credibility, argument ambiguity, and task importance on attitude judgment.
Journal of Personality and Social Psychology, 66, 460-473.
Chandon, P., Hutchinson, J. W., Bradlow, E. T., & Young, S. (2007). Measuring the value of
point-of-purchase marketing with commercial eye-tracking-data. In M. Wedel, & R. Pieters
(Eds.) Visual Marketing: From Attention to Action. Mahwah, NJ: Lawrence Erlbaum.
Christenfeld, N. (1995). Choices from identical options. Psychological Science, 6, 50-54.
Coombs, C. H., & Avrunin, G. S. (1977). Single-peaked functions and the theory of preference.
Psychological Review, 84(2), 216-30.
43
Drèze, X., Hoch S. J., & Purk, M. E. (1994). Shelf management and space elasticity. Journal of
Retailing, 70(Winter), 301-326.
Ducrot, S., & Pynte, J. (2002). What determines the eyes’ landing position in words. Perception
and Psychophysics, 64(7), 1130-1143.
Fazio, R., (1990). Multiple processes by which attitudes guide behavior: The MODE model as an
integrative framework. In M. Zanna (Ed.) Advances in Experimental Social Psychology, vol.
23, (pp. 75-109). San Diego, CA: Academic Press.
Feldman, J. M., & Lynch J. G. (1988). Self-generated validity and other effects of measurement
on belief, attitude, intention and behavior. Journal of Applied Psychology, 73, 421-435
Fiske, S.T., & Neuberg, S. L. (1990). A continuum of impression formation, from category-based
to individuating processes: Influences of information and motivation on attention and
interpretation. In M. P. Zanna (Ed.), Advances in Experimental Social Psychology. vol. 23,
(pp. 1-74). New York: Academic Press.
Friestad, M., & Wright, P. (1994). The persuasion knowledge model: People cope with
persuasion attempts. Journal of Consumer Research, 21, 1-31.
Grice, H. P. (1975). Logic and conversation. In Cole, P. & J. L. Morgan (Eds.). Syntax and
Semantics: Speech Acts, vol. 3. New York: Academic
Inman, J. J., McAlister, L., & Hoyer, W. D. (1990). Promotion signal: Proxy for a price cut? Journal
of Consumer Research, 17(June), 74-81.
Janiszewski, C. (1990). The influence of print advertisement organization on affect toward a
brand name. Journal of Consumer Research, 17(June), 53-65.
Kirmani, A., & Wright, P. (1989). Money Talks: Perceived Advertising Expense and Expected
Product Quality. Journal of Consumer Research. 16(December), 344-353.
Markus, H.R., & Kitayama, S. (1991). Culture and the self: Implications for cognition, emotion,
and motivation. Psychological Review, 98, 224-253.
44
McArthur, L. Z., & Post, D. L. (1977). Figural emphasis and person perception. Journal of
Experimental Social Psychology, 13, 520-535.
Menon, G., Raghubir, P. & Schwarz, N. (1995). Behavioral frequency judgments: An
accessibility-diagnosticity framework. Journal of Consumer Research, 22 (September), 212-
228.
Nisbett, R. E., & Wilson, T. D. (1977). Telling more than we can know: Verbal reports on mental
processes. Psychological Review, 84(3), 231-259.
Nowlis, S., & Simonson, I. (1997). Attribute-task compatibility as a determinant of consumer
preference reversals. Journal of Marketing Research, 34, 205-218.
Pham, M., & Johar, G. (1997). Contingent Processes of Source Identification. Journal of
Consumer Research, 24(4), 249-65.
Raghubir, P., & Valenzuela, A. (2006). Centre of inattention: Position biases in decision making.
Organizational Behavior and Human Decision Processes, 99(1), 66-80.
Schuman, H., & Presser, S. (1996). Questions and Answers in Attitude Surveys: Experiments on
Question Form, Wording, and Context, Sage Publications: Thousand Oaks, California.
Schwarz, N., Hippler, H., Deutsch, B., & Fritz, S. (1985). Response categories: Effects on
behavioral reports and comparative judgments. Public Opinion Quarterly, 49 (Fall), 388-395.
Senecal, S., & Nantel, J. (2004).The influence of online product recommendations on
consumers’ online choices. Journal of Retailing, 80(2), 159-169.
Shaw, J. I. (2000). Centrality preferences in choices among similar options. Journal of General
Psychology, 127(2), 157-164.
Snyder, M. (1982). When believing means doing: Creating links between attitudes and behavior.
In M. Zanna, E. Higgins, & C. Herman (Eds.) Consistency in Social Behavior: The Ontario
Symposium, vol 2 (pp. 105-130). Hillsdale, NJ: Lawrence Erlbaum.
Simonson, I. (2005). Determinants of customers’ responses to customized offers: Conceptual
framework and research propositions. Journal of Marketing, 69(1), 32-45.
45
________, & Tversky, A. (1992). Choice in context: Tradeoff contrast and extremeness aversion.
Journal of Marketing Research, 29(3), 281-295.
Sudman, S., Bradburn, N. M., & Schwarz, N. (1996). Thinking about Answers: The Application of
Cognitive Processes to Survey Methodology. San Francisco, CA: Jossey-Bass Inc.
Taylor, S. E. & Fiske, S. T. (1975). Point of view and perceptions of causality. Journal of
Personality and Social Psychology, 32, 439-445.
Wernerfelt, B. (1995). A rational reconstruction of the compromise effect: Using market data to
infer utilities. Journal of Consumer Research, 21(4), 627-33.
Wright, P. (2002). Marketplace metacognition and social intelligence. Journal of Consumer
Research, 28(4), 677-682.
46
Table 1: Results by Measure -- Study 3
Dependent
Variable1
Extreme
Left
Middle-
Left
Middle
Position
Middle-
Right
Extreme
Right
Position Effect F(4,
132)2
Rank 3.45 3.04 2.57 2.68 3.26 8.01
Intention 3.36 3.77 4.34 3.83 3.51 8.33
Attention 3.95 3.96 4.01 5.54 3.69 74.98
Popularity 3.83 4.57 5.19 4.98 4.47 16.07
Allocation 17% 20% 22% 22% 19% 7.06
Mkt. Share 16% 21% 22% 22% 19% 9.98
1: Highest rank = 1, lowest rank = 5 (Lower means represent higher preferences); intentions,
attention and popularity were all elicited on 7-point scales with higher numbers indicating higher
levels of each; allocation and market share were elicited using an open-ended format with
numbers across the five options required to add up to 100.
2: p < .05 for all.
47
Table 2: Results by Measure -- Study 5
Dependent
Variable1
Extreme
Left
Middle-
Left
Middle
Position
Middle-
Right
Extreme
Right
Self/Other by
Position
Interaction
Rank
Other
Self
17.6
19.5
19.6
20.1
27.0
18.2
14.9
21.4
20.9
19.5
χ
2 = 4.772
Purchase Intentions
Other
Self
4.00
3.62
4.04
3.69
4.51
3.91
4.01
3.87
4.22
3.91
F(1,149) = 4.772
Liking
Other
Self
4.11
3.83
4.22
3.52
4.46
3.77
3.93
3.95
4.27
3.92
F(1,149) = 5.522
Popularity
Other
Self
4.46
4.33
4.23
4.46
4.74
4.55
4.58
4.51
4.46
4.33
F(1,149) = .22
Position Factor - Quadratic
F(1, 149) = 4.72, p<.05
Allocation
Other
Self
19.82
19.08
19.94
20.02
21.69
21.45
20.57
20.67
18.32
19.96
F(1,149) = .38
Position Factor - Quadratic
F(1, 149) = 2.70, p<.10
Market Share
Other
Self
19.97
19.69
20.19
20.25
21.57
19.84
20.43
21.18
17.91
19.36
F(1,149) = .17
Position Factor - Quadratic
F(1, 149) = 3.01, p<.10
Attention
Other
Self
4.09
3.58
3.93
3.91
4.03
3.91
3.99
3.95
4.01
4.07
F(1,149) = .63
Attractiveness
Other
Self
4.22
3.84
4.03
4.13
4.30
3.97
4.15
4.23
4.26
4.25
F(1,149) = .55
Vividness
Other
Self
2.85
2.80
3.17
2.93
3.01
3.09
3.07
3.09
3.06
3.04
F(1,149) = .09
1: Rank represents the frequency of choice by position in the first two choices.
2: p < .05
48
Table 3: Results by Measure -- Study 6
Dependent
Variable
Extreme
Left
Middle
Position
Extreme
Right
Order of Elicitation by
Position
Interaction Efffect1
Choice
Ratings Before Choice
Ratings After Choice
45.3
24.7
26.7
44.7
27.9
30.1
χ
2 (2) = 6.55
Evaluation Index
Ratings Before Choice
Ratings After Choice
5.42
5.20
5.06
5.36
5.17
5.10
F(1,171) = 5.14
Quadratic Contrast
1: p < .05 for all.
49
