ArticlePDF Available

Indifference versus Ambivalence: The Effect of a Neutral Point on Consumer Attitude and Preference Measurement

Authors:

Abstract and Figures

In many situations, consumers do not have either a positive or negative attitude, and thus express a neutral response. It is hypothesized that a neutral response on a bipolar scale is caused by either (1) indifference, which is a truly neutral response, or (2) ambivalence, which is a consequence of conflict. When consumers are experiencing ambivalence, the removal of a neutral response option in an attitude or preference scale is hypothesized to lead to systematic effects on the distribution of responses. In particular, over a series of five studies it is found that when consumers are ambivalent or experience conflict due to the absence of a neutral response category, they are more likely to (1) favor the more important attribute (brand name or quality over price), (2) avoid risky options, and (3) favor a status quo option. The need for cognition moderates these effects, in that those low on need for cognition are more likely to be influenced by a neutral point. In addition, when ambiv...
Content may be subject to copyright.
Indifference versus Ambivalence: The Effect of a Neutral Point on Consumer Attitude and
Preference Measurement
September 2000
Stephen M. Nowlis, Barbara E. Kahn, and Ravi Dhar*
______________________________________________________________________________
*Stephen M. Nowlis is an associate professor of marketing in the College of Business, Arizona
State University. Barbara E. Kahn is The Dorothy Silberberg Professor of Marketing at The
Wharton School, University of Pennsylvania. Ravi Dhar is an associate professor of marketing
in the School of Management, Yale University. The authors are listed in reverse alphabetical
order and contributed equally to the research. This paper has benefited from the comments of
Mary Frances Luce and Nathan Novemsky.
Indifference versus Ambivalence: The Effect of a Neutral Point on Consumer Attitude and
Preference Measurement
Abstract
In many situations, consumers do not have either a positive or negative attitude, and thus
express a neutral response. It is hypothesized that a neutral response on a bipolar scale is caused
by either (1) indifference, which is a truly neutral response, or (2) ambivalence, which is a
consequence of conflict. When consumers are experiencing ambivalence, the removal of a
neutral response option in an attitude or preference scale is hypothesized to lead to systematic
effects on the distribution of responses. In particular, over a series of five studies it is found that
when consumers are ambivalent or experience conflict due to the absence of a neutral response
category, they are more likely to (1) favor the more important attribute (brand name or quality
over price), (2) avoid risky options, and (3) favor a status quo option. The need for cognition
moderates these effects, in that those low on need for cognition are more likely to be influenced
by a neutral point. In addition, when ambivalent consumers can be removed through an initial
task, the neutral point no longer exerts a systematic effect on the distribution of preferences. The
theoretical and measurement implications of the results are discussed
1
Measuring attitudes or preferences is an essential part of studying consumer behavior, as
consumers have attitudes toward a wide range of things. For example, consumers have attitudes
towards specific products, such as the degree of liking for detergent brands, to attitudes about
more general consumption-related behaviors, such as how often clothes should be washed (e.g.,
Solomon 1999). These judgments often influence consumer perceptions, guide purchase
intentions, and affect consumer satisfaction.
Probably the most common way to measure attitudes is through self-reports in which
people are asked to indicate their beliefs or feelings towards an object or class of objects on some
sort of scale, such as Likert (or summated ratings) scales, semantic differential scales, or
itemized ratings scales (Lehmann, Gupta, and Steckel 1998). One of the issues in designing
these ratings scales is the determination of the number of categories that should be used and
whether or not an odd or an even number of categories is appropriate. In this article we examine
the effect on attitudinal responses of including a middle response alternative (hereafter referred
to as a “neutral position”) in a scale. Specifically, building on the notion that the selection of a
neutral position in a bipolar scale often reflects attitude ambivalence, the exclusion of a neutral
position may shift the relative distribution of responses by compelling respondents to state an
opinion. Further, we propose and empirically verify that in some identifiable circumstances the
shift in the distribution of opinions due to the removal of the central point will occur in a
reasonably biased, predictable manner. Such an effect has important implications for consumer
research, suggesting that the attitude measurements can be systematically manipulated by
changing the scale that is used.
In some circumstances, consumers select the middle response option because they feel
indifferent or do not have an opinion. In these cases, as opposed to situations when consumers
2
are feeling ambivalent or conflicted, the removal of the neutral category on an attitude response
scale should not change the distribution of opinions. Conversely, if consumers are ambivalent or
experiencing psychological distress about the object (Luce, Bettman, Payne 2000) -- such as
might be the case when respondents are asked to form judgments that force them to engage in
difficult tradeoffs, or embrace risk, or move away from the status quo -- then removal of the
neutral position may significantly change the distribution of opinions. Consistent with this
premise, we show that if the respondents who are most ambivalent are filtered out through an
initial question, the distribution of responses across the two scales is similar.
In the remainder of the paper, we first review prior research relevant to the effect of a
neutral position on attitude response. This analysis leads to several hypotheses that are tested
across a series of five studies, which compare the distribution of opinion when there is a middle
option to the distribution of opinion when the middle option is not present. We report the
findings and conclude with a discussion of their implications for consumer research.
Attitude Measurement Literature Investigating Middle Response Options
Interest in the impact of a neutral point on a respondent’s judgments has been studied in
the context of survey research (Krosnick 2000) and marketing research (Lehmann et al. 1998).
In the survey and polling literatures, one area of research has focused on identifying the
antecedents that lead to a selection of a “middle response alternative.” For example, it has been
shown that people are more likely to select a middle response alternative on an issue when it is
explicitly offered to them as opposed to allowing them to volunteer that information
spontaneously (Kalton, Roberts, and Holt 1980; Schumann and Presser 1981). Respondents are
also more likely to select a middle response alternative if it is merely mentioned in the preface of
a question and not ever explicitly offered (Bishop 1987), and they are more likely to use the
3
middle option if they are uninvolved with the issue (Bishop 1990). Finally, the likelihood that
respondents will select a middle response if it is offered differs by culture (Si and Cullen 1998).
A second area has focused on whether the inclusion of a no-opinion category improves
the quality of data obtained by an attitude measure. If the exclusion of a middle point or a no-
opinion response causes respondents to report non-neutral attitudes because the survey
instrument forced them to do so, the accuracy of the data may be improved by using a scale that
includes a “no opinion” category (Hawkins and Coney 1981, Schneider 1985), and reliability and
external validity may be increased when a neutral category is offered (Buchanan and Henderson
1992). However, the evidence in favor of using a “no opinion” response as an effective filtering
device has been mixed (Krosnick 2000).
A somewhat less examined area is the consequence of including a neutral position on
interpretation of the data, or the conclusions drawn from the data. Some marketing researchers
point out that differences in results between studies using odd or even numbered categories are
not often found (e.g., Lehmann et al. 1998). Consequently, this issue has not received that much
attention in this literature, except for studies about how the data from a neutral position should be
analyzed (e.g., Ross 1969). These researchers implicitly assume that if a middle position is
omitted, the respondents forced to choose a more polar response would contribute to some form
of random error to the distribution (Klopfer 1980; Presser and Schumann 1980).
In the survey literature, there is mixed evidence as to whether or not removal of a middle
option systematically affects the distribution of responses for the other categories. Consistent
with the point made in marketing (Lehmann et al. 1998), polling/survey research has failed to
reject the null hypothesis that there is no systematic differences in univariate distributions once
middle responses are excluded (e.g., Andrews 1984; Presser and Schumann 1980). However,
4
there is some limited evidence that whether or not a middle response alternative is offered does
affect answers to polling attitude questions (Bishop 1987). One limitation of this research is that
it was primarily based on field settings and not guided by psychological principles that highlight
the process by which, and the conditions under which, a neutral position will shift the
distribution of responses. Furthermore, these surveys offered additional response categories such
as “don’t know” that are potentially substitutable with the middle response, thus making the
findings difficult to interpret and open to alternative explanations (Glucksberg and McCloskey
1981).
Reasons for Neutral Ratings: Indifference versus Ambivalence
In order to understand why a neutral position may shift the distribution of responses, we
first consider why this category is likely to be chosen. As discussed above, one reason for
selecting the midpoint of a bipolar scale may occur when the respondent truly has a neutral
attitude towards the object. A second reason for selecting the neutral position occurs if the
respondent has ambivalent feelings towards the object. For example, if a stimulus has significant
positive and negative aspects it is difficult to select a single positive (negative) response
(Krosnick 2000), and may also elicit a comparable response on a bipolar scale (Cacioppo,
Gardner, and Berntson 1997). Although past researchers (Kaplan 1972; Presser and Schumann
1980) have noted the distinction between those who choose the neutral position because they are
indifferent (i.e., neither good nor bad) or are ambivalent (i.e., both good and bad), the
consequence of this distinction has not been explored systematically until recently.
In a recent study, Larsen, McGraw and Mellers (1999) directly tested in the context of
gambles whether neutral ratings were better characterized as indifference (as is usually assumed)
or ambivalence (as they hypothesized). In their experiments, respondents indicated that they felt
5
positively toward winning but felt neutrally toward disappointing wins (wins that could have
been better). Similarly, respondents indicated that they felt negatively towards losing but
neutrally towards relieving losses (losses that could have been worse). However, when the
neutral ratings were further tested, it turned out the respondents felt both positively and
negatively in these disappointing wins and relieving losses conditions, rather than neither
positively nor negatively. Thus, rather than feeling indifferent, these neutral ratings of gambles
actually indicated conflicting feelings on the part of the respondents.
If a neutral and an ambivalent attitude towards an object map onto the same evaluation on
a bipolar scale, how does the number of categories effect the responses? We believe that if
respondents feel no conflict characterized by low activation of positive and negative aspects
towards an object, then removal of the neutral category on an attitude response scale will
contribute to some random error but will not change the distribution of responses. However, if
respondents are ambivalent or experience high activation of positive and negative thoughts about
the object, then removal of the central point will significantly change the distribution of opinions,
and further the shift will be in a biased, predictable manner. When respondents feel ambivalent
they experience conflict that can result in task-related negative emotion. We believe that
consumers’ likely coping strategies to deal with this task-related negative emotion may yield
predictable reactions to the presence or absence of a central point on a response scale. In order
to formulate these hypotheses, we first briefly review the literature on task-related negative
emotion.
Task-Related Negative Emotion
Luce, Bettman and Payne (2000) argue that choices that force respondents to forego some
attractive benefits in favor of others generate negative task-related emotion. Consumers may
6
experience this task-related negative emotion due to the actual task of thinking about the
tradeoffs involved in the judgment (Bettman, Luce and Payne 1998) or because of fear of
unfavorable evaluations from others resulting from the ultimate decision (Janis and Mann 1977).
In order to deal with this task-related emotion, consumers sometimes respond by processing
simply or not at all (Bodenhausen 1993; Keinan 1987), and more generally may seek to avoid the
relevant choice or judgment tasks (Luce et al. 2000). Such a strategy is likely to be constrained
by task related factors, such that consumers will choose the most salient avoidance mechanisms
offered by the preference elicitation task (Luce et al. 2000). In a similar vein, the positive and
negative aspects that are highlighted in a judgment task can generate negative task-related
emotion. Further, the inclusion of a neutral position may function as a salient coping mechanism
encouraging simplified processing and allowing avoidance of committing to any position. In the
absence of a neutral position, respondents are forced to confront the explicit tradeoffs that are
highlighted in arriving at an overall impression.
In contrast to the effort minimizing approach to ambivalent judgments described above,
an alternative strategy on the part of the respondent may be to engage in more effort to resolve
the conflict. This typically occurs when the cognitive ability that is required to process the
information exceeds the motivation to do so (Luce et al. 2000). In the effort engaging approach,
it is assumed that respondents carefully process information and consequently select the neutral
position to minimize negative affect. For example, Dhar (1997) showed that if individuals
expressed a similar number of favorable thoughts about each alternative in a choice set, they
were most likely to choose the “no-choice” option, that is choose to decide not to choose. In
sum, both the effort minimizing approach and effort engaging approach suggest that the neutral
position is likely to be selected by respondents in order to minimize negative affect associated
7
with ambivalent judgments. On the other hand, the absence of a neutral position makes strict
avoidance not possible and consumers are forced to confront the ambivalence and indicate an
attitude. In such instances, decision makers may search for alternative avoidance strategies.
These strategies may include choosing the option that is best on the most emotion-laden
attribute, such as safety or quality (Luce, Bettman and Payne 1997), or the most easily justifiable
dimension, generally the most important one (Simonson 1989; Slovic 1975) since thinking about
trading off an important attribute is potentially associated with increased threat. This is termed
“secondary avoidance” as these types of strategies allow respondents to avoid making the
explicit tradeoff themselves by relying on simplified heuristics that may be associated with a
lower likelihood of regret and error.
Empirical Tests About the Effect of a Middle Response on Judgments
If consumers are asked to make judgments or to indicate preferences or attitudes that they
feel ambivalent about, and thus feel task-related negative emotion at the prospect of trading off
one aspect for another, then a commitment is likely to be avoided. If a neutral or middle option
does exist, consumers will avoid commitment by choosing that option. However, if no middle
option exists, they may choose the responses in the direction that offers secondary avoidance. In
particular, they may choose in favor of the most emotion-laden attribute (e.g., choose quality
over price; Luce, et al. 2000), choose the most justifiable position (Simonson 1989), choose
against taking risks (Fishoff, Bostrom and Quadrel 1993; Viscusi, Magat and Huber 1987) or
choose in favor of the status quo (Luce 1998).
8
Study 1: Effects of a Neutral Position on Attitudinal Response
In Study 1, we test the basic hypothesis that an attitude towards an ambivalent object is
more likely to result in significant effects of the presence or absence of a middle option than
when respondents are asked to form an attitude towards an object that does not induce
ambivalence. In general, objects that have average values on all dimensions (Dhar and
Simonson 2000) are not likely to result in ambivalence due to the absence of strong positive or
negative features. In addition, choices of all-average options are less likely to be negatively
evaluated by others (Simonson and Nowlis 2000) so they are less likely to cause task-related
distress. In contrast, items that are more extreme and similar on good and bad attributes
(hereafter called “extreme” options) are likely to cause ambivalence and fear of negative
evaluation, and thus are more likely to cause negative task-related emotion (Shafir 1993;
Thompson, Zanna, and Griffin 1995). Specifically, such options are likely to result in a strong
tendency to both approach and avoid the same object. Hence we hypothesize:
H1a: Due to anticipated attitude ambivalence, there will be significant differences in the
distributions of opinion for an extreme option using an even point scale as compared to using
an odd point scale (with a neutral point). No such differences in the distribution of opinion
will exist for an all-average option.
Further, when forced to express an attitude for an extreme option, those who feel
ambivalent are more likely to favor the most important attribute, generally quality versus price or
other costs. This premise has received some support from the choice literature. For example,
Simonson (1992) showed that consumers who considered the possibility of regret resulting from
the choice of a wrong option were more likely to choose a high price, high quality option over a
low price, low quality option. Similarly, Simonson and Tversky (1992) showed a systematic
bias in favor of high quality, high price options. In addition, Shafir (1993) demonstrated a
9
greater focus on the positive attributes of the extreme option in a choice task. Hence, we
hypothesize:
H1b: When forced to express an attitude toward an extreme option in the case of an even point
scale, respondents will favor the more important dimension, e.g., quality (over costs).
Method
Subjects were 68 undergraduate marketing students who completed the paper and pencil
questionnaires as part of a class requirement for the introductory marketing course. Each subject
indicated their attitudes for an all-average and an extreme product in three separate product
categories (restaurants, calculator, and personal computers). For example, with restaurants,
subjects evaluated an all-average option with: “average quality (2.5 stars), average wait, average
selection, and average atmosphere.” They also evaluated an extreme option with: “high quality
(4 stars), long wait, wide selection, and dull atmosphere.” Each option was evaluated on a
separate page of the survey. The questions for calculators and personal computers were similar in
structure to those for restaurants.
1
The order of preference ratings was counter-balanced across
respondents.
There were two between-subjects conditions. In one condition, subjects used a 4-pt
attitude scale, with each of the four points labeled: 1= dislike very much, 2=dislike somewhat,
1
We constructed the alternatives so that the extreme options would be superior on the more important
attributes, since H1b predicts that subjects will be more likely to choose the extreme options when forced
to do so. To do that, we had a separate group of 30 student subjects rate each attribute of each alternative
by answering the following question: “How important is (attribute) in deciding how much you like or
dislike (option),” and they responded on a seven-point scale anchored from “Not at all important” to
“Very important.” We found that subjects rated certain attributes as more important than others (p < 0.05
for all cases), and these attributes were the ones favored when subjects were forced to decide (H1b). For
example, subjects rated reliability and warranty as more important than price and function keys for
calculators (and then gave a more favorable rating to the extreme calculator, which was better on the more
important attributes, when forced to do so). We tested the attributes of the alternatives used in Study 4 in
a similar manner. There we constructed choice sets such that the quality attributes were found to be more
important than the cost or price attributes.
10
3=like somewhat and 4=like very much. In the second condition, subjects used a 5-pt scale that
also included a neutral point, labeled as such.
Results
To test the hypotheses, we rescaled the data into a form where direct comparisons could
be made across conditions. We followed other research by rescaling the responses in the odd
point scale by dropping the responses to the neutral point, leaving the remaining 4 points, which
we then compared directly to the same four points in the even point scale (e.g., Bishop 1987;
Presser and Schumann 1980)
2
. For example, in Table 1, looking at the average across the three
tested categories, 33% liked the extreme options, while 68%
3
did not, after rescaling these values
when the neutral point was removed. Without removing the neutral point, 26% liked the extreme
options, 55% did not, and 18% chose the middle response. Thus, Table 1 shows the rescaled
values for the odd point scale, but also shows the percentage of respondents selecting the neutral
point so that we have presented the complete results.
H1a predicts that there will be a significant effect of the scale on the distribution of
responses for extreme options, but no difference for all-average options. Table 1 shows,
averaged across the three categories, that 33% liked the extreme options when using the odd
point scale. When using the even point scale, 52% preferred the extreme options (which were
better on the more important attributes), for an increase of 19%. For the average options, 66%
liked these alternatives with both the even and odd point scales, and thus there was no change.
We tested H1a and H1b with a logistic regression model, where the responses were modeled as a
2
The hypotheses can also be tested in a different manner. Specifically, we can test the average value of
the response across the scale manipulation. For example, a response of “like very much” would receive a
score of “2”, a response of “neutral” would receive a score of “0”, and a response of “dislike somewhat”
would receive a score of “-1”, etc. We then can average the scores across either the even or odd point
scales, and compare the scores with a t-test. When we analyzed the results using this method, we found
similar results. Thus, for brevity these results are not reported here, but are available for the interested
reader from the authors.
11
function of the following independent dummy variables: (1) a variable indicating whether the
odd or even scale was used, (2) a variable indicating whether subjects evaluated extreme or
average options, (3) a two-way interaction between these variables, which tests H1a, (4) a two-
way interaction between the scale manipulation and the all-average options, and (5) a two-way
interaction between the scale manipulation and the extreme options. The last two interactions
test to see if there are significant differences across the tested categories (e.g., Chernev 1997;
Dhar 1997). H1b was tested with coefficient (1) above when looking only at the response to the
extreme options.
First, we found that the interaction between coefficients (1) and (2) was significant
(
2
? (1) = 5.51, p < 0.05), supporting H1a. Next, H1b was supported as there was a significant
effect of the scale manipulation for the extreme options (
2
? (1) = 8.25, p < 0.01). Finally, we
found that there were no significant differences in the effects across the categories, for both all-
average (
2
? (2) = 0.24, ns) and extreme options (
2
? (2) = 1.2, ns). Thus, the results were
consistent across the three categories.
Study 2: Effects of a Neutral Position on Attitudes toward Risky Options
Study 1 showed that if the judgment context evoked ambivalence, such as in the case of
an object eliciting positive and negative thoughts, then an even versus an odd point rating scale
can affect the distributions of opinion in a predictable manner. In Study 2, we show that risky
options with mixed outcomes can also lead to situations where some respondents are looking to
avoid making judgments, and therefore once again there will be systematic differences in
preferences depending upon whether an even or odd point scale is used.
3
Numbers do not add to 100% due to rounding error.
12
Consistent with the coping strategy described above that consumers use to avoid making
difficult tradeoffs, consumers may rush through the decision process or simply not process at all
(if possible) to avoid negative, risk-associated feelings that are induced by being asked to form
preferences for risky options (Bauer 1960). Risk-related tradeoffs are generally difficult and
potentially threatening, and consumers have been found to have difficulty weighing even small
risk increases against benefits (Fischhoff, Bostrom, and Quadrel 1993). Thus, if given the
chance to avoid commitment by choosing a neutral category, those who find themselves
ambivalent about trading off the risk for the benefit will take that option. However, if they are
forced to express an evaluation, then they are likely to opt for the lesser emotion-laden attribute,
which would be to opt for security, i.e., the more risk averse alternative. This is consistent with
past data that has shown that consumers are generally averse to accepting increases in risk
(Viscusi, Magat, and Huber 1987), and that increases in risk are generally considered more
negative than the corresponding decreases in risk are considered positive. Hence, we
hypothesize:
H2: Due to anticipated attitude ambivalence, there will be significant differences in the
distributions of opinion for a risky option using an even point scale as compared to using an
odd point scale (with a neutral point). Further, consumers will give a lower rating to a risky
investment with an even-point scale than with an odd-point scale (with a neutral point).
Method
Subjects were 193 undergraduate marketing students who completed the paper and pencil
questionnaires as part of a class requirement for the introductory marketing course. Subjects
indicated the attractiveness for three different investment opportunities ($100, $300, and $500).
Each investment involved the possibility of a gain and a loss. Given the anticipated risk
aversion, the options were constructed so that the gain was higher than the loss, so that the
expected value of the gamble was higher than the expected value of doing nothing. For instance,
13
the $100 investment was described as: “Initially invest $100. After 3 months, there is a 50%
chance this investment will be worth $150. But, after 3 months, there is a 50% chance this
investment will only be worth $75.” For the $300 investment, there was a 50% chance of the
investment being worth $500, and a 50% chance of the investment being worth $200. For the
$500 investment, we not only increased the amount of the gain, but also increased the likelihood
that the gain would occur by indicating a 75% chance of the investment being worth $700 and a
25% chance of the investment being worth only $200. The order of the gambles was counter-
balanced across respondents.
As in Study 1, there were two between-subjects conditions. In one case, subjects
evaluated each investment on a 4-pt scale, with 1=very unattractive, 2=somewhat unattractive,
3=somewhat attractive and 4=very attractive. In the other condition, we also included a neutral
point labeled as such.
Results
As we did in Study 1, we analyzed the results for this study by removing the responses to
the neutral point in the 5 point scale, and then comparing responses directly to the 4 point scale.
H2 predicts that consumers will give a lower rating to a risky investment with an even-point than
with an odd-point scale. Table 2 shows that, averaged across the three problems, 66% found the
investments to be attractive when using the odd point scale, and 55% found the investment to be
attractive when using the even point scale. We tested H2 with a logistic regression model, where
the responses were modeled as a function of the following independent dummy variables: (1) a
variable indicating whether the odd or even scale was used, which tests H2 and (2) a two-way
interaction between the scale manipulation and the three investments, which tests whether the
effects were different across the problems. Supporting H2, we found the effect of the scale
14
manipulation to be significant (
2
? (1) = 9.16, p < 0.01). We also found that there were no
significant differences across the individual investments (
2
? (2) = 0.45, ns).
Study 3: Effects of a Neutral Position on Attitudes Toward a Status Quo Object
As we have suggested so far, if respondents can avoid resolving their ambivalence by
using a neutral rating point, they will. However, if that neutral rating point does not exist, such
as in an even point scale, then they will look for another way to avoid “making waves,” by
choosing the status quo option, or keeping things the way they are. This finding is consistent
with laboratory work on decisions involving tradeoff difficulty. In two experiments, Luce
(1998) shows that the choice of status quo alternatives mitigates the negative task-related
emotion, and that the more conflicted consumers feel about the decision, the more likely they
will choose the status quo option given that it is available. Thus the choice of the status quo
option in an even-point scale is somewhat equivalent to the choice a neutral point in an odd point
scale. Hence, we hypothesize:
H3: Due to anticipated attitude ambivalence in a choice that involves choosing a new option or
remaining with the status quo, the distribution of response favors the status quo option when
the scale is even point as compared to an odd point scale.
Method
Subjects were 244 undergraduate marketing students. We conducted a 2 (item is status
quo option or not) X 2 (4 or 5 point scale) between-subjects design. Respondents decided
between two options, one of which was labeled as their “current” option (see Figure 1). For
example, when deciding between apartments, respondents were asked: “Imagine that you have
been renting a 1-bedroom apartment (Current Apartment below). Your current lease is up and
you have the chance to stay in your current apartment or move into a different apartment (New
Apartment below). What would you do?” As mentioned above, we rotated whether the first or
15
second apartment was designated as the status quo (Current) option. Subjects then either decided
between the two options on a 4-point or 5-point scale, as in Study 1.
Unlike in the other studies described so far, in this study the neutral point really does not
equal indifference. Subjects who choose the neutral point here are implicitly choosing to stay
with the status quo. Therefore support for the hypothesis here will also provide evidence that
subjects understand that a neutral point is not just an option to not process information but rather
is, in this case, a conscious vote in and of itself. Thus, as a corollary to show internal validation
for our methods, the percentages of subjects who choose the middle point on the 5-point scale
should be approximately equal to the advantage that the status quo option receives over the new
option on the 4-point scale in this between subjects design.
Results
As with the first two studies, we analyzed the results for this study by removing the
responses to the neutral point in the 5 point scale, and then comparing responses directly to the 4
point scale. H3 predicts that the status quo option will be favored when responses are made on
the even-point scale. Table 3 shows that, averaged across the three problems, 52% preferred the
status quo option when using the odd point scale, compared to 61% with the even point scale.
We tested H3 with a logistic regression model, where the responses were modeled as a function
of the following independent dummy variables: (1) a variable indicating whether the odd or
even scale was used, (2) a variable indicating whether option A or option B was the status quo
item, and (3) a two-way interaction between the scale manipulation and the three tested
categories. We found that the coefficient testing the scale manipulation was significant (
2
? (1) =
7.03, p < 0.01), supporting H3. In addition, the coefficient testing whether option A or option B
was the status quo item was not significant (
2
? (1) = 1.29, ns), indicating that it did not matter
which of the particular options was used as the status quo item. Finally, we found that there
16
were no significant differences in the effects across the categories (
2
? (2) = 0.88, ns). Our
corollary here is also supported. In our between-subjects design, 9% of the subjects checked the
middle option on the 5-point scale and that is equal to the advantage that the status quo option
received on the 4-point scale.
Study 4: Effects of a Neutral Position on the Preference For Price/Quality Tradeoffs
As we have shown now in three studies, objects that induce ambivalence or some task-
related negative emotion are more likely to reveal significant differences depending upon
whether or not an even or odd scale is used to record attitudes. Although our focus in the studies
so far was on conflict induced by a single object that highlighted positive and negative aspects,
conflict can also arise when the decision involves a choice between two attractive goals, where
none has a systematic advantage. Surveys often use bipolar scales to measure the relative
strength of preference among pairs of options. In the next study, we extend this proposition by
having respondents indicate their preferences for a high quality, high price option versus a low
price, low quality option. Furthermore, we provide some evidence for our hypothesized process.
When asked to indicate preferences for high quality/high price options versus low
price/low quality options, we hypothesize that as in the previous three experiments, some
respondents will feel ambivalence with respect to the tradeoffs necessary to form this preference.
Hence there will be a significant difference in the distribution of opinions across an odd and an
even scale. Further, we hypothesize that when these ambivalent respondents are forced to
express a preference in one direction in an even-point scale, they will move towards the high
quality, high price option. There are several reasons for this prediction. First, Luce et al. (2000)
argue that quality attributes are more likely to elicit a wider range of emotional levels because
they are associated with a wider range of goals. Price, on the other hand, is associated with one
17
goal, which is saving money. The fungibility of price attributes should also reduce anticipated
regret associated with losses on price. As mentioned earlier, both Simonson (1992) and
Simonson and Tversky (1992) showed that consumers exhibited systematic biases in favor of
high quality/high price options at the expense of low quality/low price options. Hence, we
hypothesize:
H4a: In a choice between a low-price, low-quality (low-tier) brand and a high-price, high-
quality (high-tier) brand, the high quality/high price brand will gain more share than the other
option when a neutral point is eliminated.
The other purpose of this experiment was to try and provide some support for the
hypothesized process driving our results. We are hypothesizing that the significant effects
between even and odd point scales occur because some subjects are experiencing ambivalence,
or task-related negative emotion and are coping with this situation by reacting heuristically. The
degree to which respondents are likely to engage in problem-focused coping is likely to differ
across individuals. Based on the need for cognition scale (Cacioppo, Petty, and Kao 1984),
consumers who are high on the need for cognition scale are more likely to cope with a decision
in a cognitive or problem-focused way (Lazarus 1991, 1999). In a similar vein, Thompson,
Zanna, and Griffin (1995) suggest that NFC is negatively related to ambivalence because
individuals high in NFC are more likely to work through or reconcile contradictory information
about attitude objects. In contrast, consumers low on the need for cognition scale are more likely
to experience ambivalence and engage in the emotion-focused coping strategies described
earlier. Accordingly, consumers that are low on the need for cognition scale are more likely to
use heuristics, such as avoidance strategies, to solve difficult tradeoffs. Hence, we hypothesize:
H4b: Need for cognition acts as a moderator. Those who are low in NFC are more likely to be
affected by the removal of the neutral point than those who are high in NFC.
18
Method
Subjects were 165 undergraduate marketing students. We conducted a 2 (4 or 5 point
scale) X 2 (high or low in NFC) between-subjects design. On each page of the questionnaire,
respondents decided between two options, one of which was a high-tier brand (higher price and
perceived quality)
4
and the other was a low-tier brand. Subjects evaluated products in three
categories: televisions, camera film, and computers. For example, when deciding between
televisions, respondents evaluated a Sony priced at $289 and an Emerson priced at $189. Each
was described as a 20-inch model with a sleep timer and a 150 channel quartz tuner. To measure
NFC, we used the 18-item scale developed by Cacioppo et al. (1984).
Results
H4a predicts that a high-tier brand will gain choice share over a low-tier brand in even
point compared to odd point scales. Table 4 shows that, averaged across the three problems,
44% were likely to buy the high-tier brands when using the even point scale, compared to 33%
when using the odd point scale. H4b predicts that this effect will be greater for those low in
NFC. To examine this hypothesis, we followed other research which has used a median split to
separate those high and low on a personality variable (e.g., Simonson and Nowlis 2000).
Consistent with H4b, we find a difference across the scale manipulation of 16% for those low in
NFC, but only a 7% difference for those high in NFC.
We tested H4a and H4b with a logistic regression model, where the responses were
modeled as a function of the following independent dummy variables: (1) a variable indicating
whether the odd or even scale was used, which tests H4a, (2) a variable indicating whether the
4
A pretest (N=40 students) was conducted in which the brands were rated in terms of (1) overall product
quality and (2) brand performance. On both scales, the low quality brands were rated significantly lower
than the high quality brands (p < 0.05 for each scale).
19
subject was high or low on NFC, (3) a two-way interaction between these variables, which tests
H4b, and (4) a two-way interaction between the scale manipulation and the three categories,
which tests whether the effects were different across the problems. Supporting H4a, we found
the effect of the scale manipulation to be significant (
2
? (1) = 8.23, p < 0.01). Supporting H4b,
we found that the two-way interaction was significant (
2
? (1) = 5.71, p < 0.05). We also found
that there were no significant differences across the individual categories (
2
? (2) = 1.88, ns).
Study 5: The Moderating Effect of a Filter That Removes Attitude Ambivalence
Our results so far suggest that when judgments about an object or between objects
involve ambivalence, the presence of a neutral point allows a respondent to avoid the negative
affect associated with such judgments. In contrast, the absence of a neutral position makes it
difficult to cope with ambivalence by avoiding commitment. These respondents use secondary
coping strategies that shift the distribution of responses in favor of certain types of alternatives.
Although ambivalence was manipulated by creating different types of choice sets, not all
respondents are likely to experience ambivalence in the same situation. If such respondents are
removed through an initial filtering task, the pattern of responses across the two types of scales
should not be significantly different. In other words, including an option that removes the
response of uncertain subjects should act as a moderator. If, according to our theory, the middle
point alters the response pattern by (i) capturing respondents who are ambivalent about their
attitudes and (ii) by reducing the degree of conflict that is experienced in a decision situation,
providing an option that first removes these respondents should reduce the skewness in
distribution between the two conditions.
H5: If respondents who are ambivalent about their attitudes are first removed from the analysis,
there will be no difference between responses on an even- versus odd-point scale. If these
respondents are included, the same biases revealed in Study 4 should be observed.
20
Method
Subjects were 272 undergraduate marketing students. We conducted a 2 (4 or 5 point
scale) X 2 (“opting out” option available or not) between-subjects design. As in Study 4,
subjects decided between a high-tier and a low-tier brand in the same three product categories:
televisions, camera film, and computers. In the condition where an opting out option was
available, subjects first decided whether or not they were ready to express a preference between
the options provided. If they were ready, they then decided between the two options. If not, they
moved to the next problem. For instance, when deciding between televisions, subjects were
asked: “First, we would like to know whether or not you feel you are ready to make a decision
between the two televisions described below.” They had the option to circle either: “(a) I am
having a difficult time deciding between the two televisions (please turn to next page),” or “(b) I
feel comfortable making a decision between the two (next answer the question below).” Directly
below this, subjects were asked: “Next, if you feel comfortable about deciding (and circled (b)
above), please circle one number below (circle a number on the right if you would be more likely
to buy the Sony, or a number on the left if you would be more likely to buy the Emerson).” In
the other condition, this option was not available, and the subjects simply decided between the
options (as was done in the first four studies).
Results
H5 predicts that removing the responses of subjects who are not ready to make a
commitment will reduce the effect of the scale on responses. Table 5 shows that in the absence
of the filtering task, 37% preferred the high-tier brands when using the odd point scale, and 47%
preferred these brands when using the even point scale. This is consistent with the results of the
prior study and with H4a. However, if we focus on the filtered respondents, 50% were likely to
21
buy the high-tier brands when using the odd point scale, compared to 49% when using the even
point scale. We tested H5 with a logistic regression model, where the responses were modeled
as a function of the following independent dummy variables: (1) a variable indicating whether
the odd or even scale was used, (2) a variable indicating whether or not an “opting out” option
was available, (3) a two-way interaction between these variables, which tests H5, and (4) a two-
way interaction between the scale manipulation and the three categories, which tests whether the
effects were different across the problems. Supporting H5, we found that the two-way
interaction was significant (
2
? (1) = 4.16, p < 0.05). We also found that there were no
significant differences across the individual categories (
2
? (2) = 2.19, ns). In order to test H4a,
we analyzed the main effect of the scale manipulation when the opting out option was not
available, and this was significant (
2
? (1) = 5.74, p < 0.05).
General Discussion
Consumer attitudes are often measured using scales that either include or exclude a
neutral position. The main purpose of the present set of studies was to examine the consequence
of having a neutral position on scales used to measure consumer attitudes and preferences.
Although previous survey research has examined the effect of including a neutral point on
responses in an ex post manner, the data revealed mixed patterns and there was no theory
explaining when and why consumers would select that neutral position. Recent research
suggests that the neutral position may be capturing responses that are ambivalent and not just
indifferent. Building on research in avoidant choices and conflict, this article suggests how the
distribution of responses changes systematically when respondents are forced to express an
opinion. Our findings allowed us to both demonstrate important effects of excluding a neutral
position on the distribution of responses and to clarify the processes that are involved in such
22
effects. Five studies examined the effect of excluding a neutral position on attitude responses by
(1) testing predictions regarding distribution of responses for different objects, chosen
strategically so that the degree of expected attitude ambivalence would vary in systematic,
predictable ways, (2) examining additional types of tradeoffs (e.g., risk, status quo) to show the
shift in strategies when a neutral point is excluded, and (3) testing the boundaries under which
the predicted effects operate by using both an individual differences scale and a filtering task.
These studies are summarized next.
We first examined the effect of including a neutral position for ambivalent stimuli. Study
1 showed that excluding a neutral position leads to expressing a more positive attitude response
for the extreme option. In contrast, the exclusion of a neutral position did not influence the
relative distribution of responses for an average option. Study 2 extended the effect of excluding
a neutral position to judgments of risky options. As predicted, respondents gave a lower
attractiveness rating to a risky investment when a neutral position was excluded. Study 3
provided support for the underlying theoretical rationale by investigating the effects of including
a neutral position on preference for a status quo option. As hypothesized, the status quo option
was more preferred when the neutral position was excluded.
The next two studies clarified the process by which the introduction of a neutral position
influences the distribution of responses. Study 4 examined the moderating effect of need for
cognition on the effect of introducing a neutral position. We predicted and found that the low
need for cognition respondents were more susceptible to the absence of a neutral position on the
scale provided. Study 5 showed that if the undecided subjects were excluded by using a filtering
question, the difference in the distribution of responses to the odd and even scales was no longer
found. In light of these findings, we discuss the theoretical and marketing implications.
23
Theoretical Implications
Note that in general the selection of the neutral position is not limited to conditions where
respondents are ambivalent. Indeed, we find that a greater percentage of respondents selected
the neutral position for the average option than the enriched option. However, only for the cases
in which the stimulus was likely to induce ambivalence did the absence of a neutral position shift
the response in the direction that minimized likelihood of regret and error. These findings thus
support the notion that different causes underlie the selection of a neutral position. Studies 4 and
5 in particular provided support for the theoretical model by demonstrating that the difference in
distribution occurred for ambivalent objects as well as filtering out such respondents through an
initial task.
The article focused on the consequence of including a neutral position on the attitudinal
responses. In addition to influencing the attitudes, the shift in distribution of responses may also
carry over to subsequent differences in intent and behavior. Previous research demonstrates that
measuring intent has a significant effect on actual behavior (Morwitz, Johnson, and Schmittlein
1993). This suggests that scale differences in measuring attitudes may subsequently result in
different levels of behavioral response. In particular, because scales that excluded a neutral
position typically involve an overweighting of features that are associated with quality,
subsequent behavior should favor high quality items.
A general implication of our findings is that including a neutral position will
systematically distort attitude response distributions when attitudes are ambivalent. However,
certain other tasks may also induce greater ambivalence, such as judgments about an object in
the future. For example, consider the task of asking respondents to give a judgment about a
candidate for an election that is several months down the road. Such a task may lead to a greater
24
distortion between even and odd point scales than for the same candidate if the election were to
be held next week. Further, our experiments were confined to scales with four or five categories.
Future research should examine whether four- and six- point scales are more similar to each
other than they are to a five-point scale. Another line of future research might look at the effect
of a neutral position when judgments are made under time pressure. For instance, time pressure
may enhance the preference for a neutral position when the stimuli evokes ambivalence and
hence further distort the distribution between odd and even point scales.
Implications for Measurement of Consumer Attitudes
The studies also have an important practical goal in pointing out new biases in
measurement. Our data clearly demonstrate ways in which consumer responses can be
significantly altered by excluding a neutral category when respondents are ambivalent. Because
few products achieve total dominance in the marketplace, the attitudes towards most objects
involve some degree of mixed feelings on the part of respondents. For example, consumer
evaluation of new fast food types that are high in calories and high on taste may evoke different
responses on odd and even point scales. Moreover, evaluation of consumer satisfaction
judgments may also be systematically different across the two types of the scales (e.g., Ryan,
Buzas, and Ramaswamy 1995).
Although our data suggests that scales that include the neutral selection category
produces a different response from the scales that exclude this category, a question that arises is
which of the two scales is likely to best reflect the underlying attitudes. The answer to this
question is further complicated by the notion that there may be no single evaluation in memory
but rather these are often constructed when required (Schwarz and Bohner 2000). Under such
circumstances, researchers may consider the particular goals before deciding which scale is
25
appropriate. For example, excluding the neutral position may be appropriate if the survey is
trying to determine the voting patterns for the likely voters as these voters have committed to
vote. Alternatively, for categories where respondents have not yet committed to act, the
inclusion of a neutral position may be more appropriate.
26
Table 1: Results from Study 1
(% choosing each response)*
Product
category
Type of
option
type of
scale
Dislike
very much
Dislike
somewhat
Like
somewhat
Like very
much
Choosing
the
Neutral
Point
Restaurants Extreme
5 pt. scale
21 46 29 4 17
4 pt. scale 7 39 25 29
Average
5 pt. scale
7 47 47 0 48
4 pt. scale 7 41 52 0
Calculators Extreme
5 pt. scale
24 43 24 10 28
4 pt. scale 14 25 32 29
Average
5 pt. scale
14 23 55 9 24
4 pt. scale 18 18 43 21
Computers Extreme
5 pt. scale
23 46 31 0 10
4 pt. scale 11 50 36 4
Average
5 pt. scale
4 9 83 4 21
4 pt. scale 0 18 61 21
Totals Extreme
5 pt. scale
23 45 28 5 18
4 pt. scale 11 38 31 21
Average
5 pt. scale
8 26 62 4 31
4 pt. scale 8 26 52 14
* We followed other research by rescaling the responses in the odd point scale by dropping the
responses to the neutral point, leaving the remaining 4 points, which we then compared directly
to the same four points in the even scale. Thus, we show the rescaled values for the odd point
scale, but also show the percentage of respondents selecting the neutral point so that we have
presented the complete results.
27
Table 2: Results from Study 2
(% choosing each response)
Investment
opportunity
type of
scale
Very
unattractive
Somewhat
unattractive
Somewhat
attractive
Very
attractive
Choosing
the
Neutral
Point
$100 – 5 pt.
scale
10 22 61 6 17
4 pt. scale 14 31 43 13
$300 – 5 pt.
Scale
10 28 52 10 18
4 pt. Scale 11 31 44 15
$500 – 5 pt.
Scale
6 25 58 11 12
4 pt. Scale 14 36 34 16
Total 5 pt.
Scale
9 25 57 9 16
4 pt. Scale 13 33 40 15
28
Table 3: Results from Study 3
(% choosing each response)
Product
category
type of scale
Much more
likely to
switch
Somewhat
more likely
to switch
Somewhat
more likely
to stay with
status quo
Much more
likely to stay
with status
quo
Choosing
the neutral
point
CD players
5 pt. scale
18 27 27 28 9
4 pt. scale 16 23 28 34
Apartments
5 pt. scale
12 29 36 22 10
4 pt. scale 18 22 32 29
Televisions
5 pt. scale
18 36 28 18 7
4 pt. scale 12 30 30 29
Total 5 pt.
scale
16 31 30 22 9
4 pt. scale 15 25 30 31
29
Table 4: Results from Study 4
(% choosing each response)
Product
category
type of scale
NFC
Much
more
likely to
buy low-
tier brand
Somewhat
more
likely to
buy low-
tier brand
Somewhat
more
likely to
buy high-
tier brand
Much
more
likely to
buy high-
tier brand
Choosing
the neutral
point
Computers
5 pt. scale
High 43 21 28 9 2
4 pt. scale 25 27 34 14
5 pt. scale Low 35 28 14 24 15
4 pt. scale 26 24 29 21
Televisions
5 pt. scale
High 50 21 18 11 8
4 pt. scale 43 16 25 16
5 pt. scale Low 36 32 13 19 9
4 pt. scale 32 16 32 21
Camera Film
5 pt. scale
High 55 14 18 14 8
4 pt. scale 30 36 14 21
5 pt. scale Low 36 39 16 10 9
4 pt. scale 26 32 24 18
Totals 5 pt.
scale
High 49 18 22 12 6
4 pt. scale 33 27 24 17
5 pt. scale Low 35 33 14 18 11
4 pt. scale 28 24 28 20
Totals 5 pt.
scale
43 25 19 14 8
4 pt. scale 31 25 26 18
30
Table 5: Results from Study 5
(% choosing each response)
Product
category
type of scale
“Opting-
out”
option
available
Much
more
likely to
buy low-
tier brand
Somewhat
more
likely to
buy low-
tier brand
Somewhat
more
likely to
buy high-
tier brand
Much
more
likely to
buy high-
tier brand
Choosing
the neutral
point
Computers
5 pt. scale
Yes 20 16 36 28 8
4 pt. scale 31 13 31 25
5 pt. scale No 20 23 30 27 7
4 pt. scale 18 19 33 30
Televisions
5 pt. scale
Yes 39 20 15 26 6
4 pt. scale 40 17 19 25
5 pt. scale No 42 31 17 11 6
4 pt. scale 30 24 30 16
Camera Film
5 pt. scale
Yes 43 10 21 25 12
4 pt. scale 44 11 25 21
5 pt. scale No 47 26 10 18 10
4 pt. scale 28 39 21 13
Totals 5 pt.
scale
Yes 35 15 24 26 9
4 pt. scale 38 14 25 24
5 pt. scale No 36 27 19 18 8
4 pt. scale 25 27 28 19
Totals 5 pt.
scale
36 21 21 22 9
4 pt. scale 32 20 27 22
31
Figure 1: Example Product Category from Study 3
Apartments
Imagine that you have been renting a 1-bedroom apartment (Current Apartment below).
Your current lease is up and you have the chance to stay in your current apartment or move into
a different apartment (New Apartment below). What would you do?
Current Apartment
New Apartment
• New apartment building
• Color TV and cable
• New wall-to-wall carpeting
• High security deposit
• Has dishwasher and refrigerator
• Cost of heat included in rent
• Nice new furniture
• High security deposit
On the scale below, please indicate your relative preference between the two options
(please circle one number below). For example, if you circle a “0”, that means that you do not
have a preference between keeping your current apartment or take the new one. Or, for instance
if you would be more likely to keep the Current Apartment, you would circle a “1” or “2” on the
left part of the scale, or if you would be more likely to take the New Apartment, you would circle
a “1” or “2” under the right part of the scale. Please circle the one number that best describes
your preference between the two options:
More Likely to Keep Current Apartment More Likely to Take New Apartment
Much More Slightly More No Preference Slightly More Much more
2 1 0 1 2
32
References
Andrews, F. M. (1984), “Construct Validity and Error Components of Survey Measures: A
Structural Modeling Approach,” Public Opinion Quarterly, 48, 409-442.
Bauer, Raymond A. (1960), “Consumer Behavior as Risk Taking,” in Proceedings of the
American Marketing Association, ed. Robert S. Hancock, Chicago: American Marketing
Association, 389-398.
Bettman, James R., Mary Frances Luce, and John W. Payne (1998), “Constructive Consumer
Choice Processes,” Journal of Consumer Research, 25 (December), 187-217.
Bishop, George F. (1987), “Experiments with the Middle Response Alternative in Survey
Questions,” Public Opinion Quarterly, 11, 220-232.
Bishop, George F. (1990), “Issue Involvement and Response Effects in Public Opinion Surveys,”
Public Opinion Quarterly, 54 (Summer), 209-218.
Bodenhausen, Galen V. (1993), “Emotions, Arousal, and Stereotypic Judgments: A Heuristic
Model of Affect and Stereotyping,” in Affect, Cognition, and Stereotyping, ed. Diane M.
Mackie and David L. Hamilton, San Diego, CA: Academic Press, 13-37.
Buchanan, Bruce and Pamela W. Henderson (1992), “Assessing the Bias of Preference Detection
and Identification Measures of Discrimination Ability in Product Design,” Marketing
Science, 11 (Winter), 64-75.
Cacioppo, John T., Wendi L. Gardner, and Gary G. Berntson (1997), “Beyond Bipolar
Conceptualizations and Measures: The Case of Attitudes and Evaluative Space,” Personality
and Social Psychology Review, 1(1), 3-25.
Cacioppo, John T., Richard E. Petty, and Chuan F. Kao (1984), “The Efficient Assessment of
33
Need for Cognition,” Journal of Personality Assessment, 48 (June), 306-307.
Chernev, Alex (1997), “The Effect of Common Features on Brand Choice: Moderating Role of
Attribute Importance,” Journal of Consumer Research, 23 (March), 304-311.
Dhar, Ravi (1997), “Consumer Preference for a No-Choice Option,” Journal of Consumer
Research, 24 (September), 215-232.
Dhar, Ravi and Itamar Simonson (2000), “The Effect of Forced Choice on Choice,” Yale
University working paper, June.
Fischhoff, Baruch, Ann Bostrom, and Marilyn J. Quadrel (1993), “Risk Perception and
Communication,” Annual Review of Public Health, 14, 183-203.
Glucksberg, S. and M. McCloskey (1981), “Decisions About Ignorance: Knowing That You
Don’t Know,” Journal of Experimental Psychology: Human Learning and Memory, 311-325.
Hawkins, D.I. and K.A. Coney (1981), “Uninformed Response Error in Survey Research,”
Journal of Marketing Research (August), 370-374.
Janis, Irving L. and Leon Mann (1977), Decision Making: A Psychological Analysis of Conflict,
Choice, and Commitment. New York: Free Press.
Kalton, G. J., Julie Robert, and D. Holt, (1980), “The Effects of Offering a Middle Response
Option with Opinion Questions,” Statistician, 29, 65-78.
Kaplan, Kalman J. (1972), “On the Ambivalence-Indifference Problem in Attitude Theory and
Measurement: A Suggested Modification of the Semantic Differential Technique,”
Psychological Bulletin, 77 (May), 361-372.
Keinan, Giora (1987), “Decision Making Under Stress: Scanning of Alternatives Under
Controllable and Uncontrollable Stress,” Journal of Personality and Social Psychology, 52
(March), 639-644.
34
Klopfer, F. (1980), “The Middle Most Choice on Attitude Items: Ambivalence, Neutrality, or
Uncertainty?,” Personality and Social Psychology Bulletin, 6, 97-101.
Krosnick, Jon A. (2000), “The Causes and Consequences of No-Opinion Responses in Surveys,”
in R. M. Groves, D.A. Dillman, J.L. Eltinge, and R. J. A. Little (Eds.), Survey nonresponse.
New York: Wiley.
Larsen, Jeff T., A. Peter McGraw and Barbara A. Mellers (1999), “Do Neutral Ratings Imply
Indifference or Ambivalence?” poster presented at the meeting for of the Society for
Judgment and Decision Making, Los Angeles, CA (November).
Lazarus, Richard S. (1999), Stress and Emotion: A New Synthesis, New York: Springer.
Lazarus, Richard S. (1991), Emotion and Adaptation, New York: Oxford University Press.
Lehmann, Donald R., Sunil Gupta and Joel H. Steckel (1998), Marketing Research, Reading,
MA: Addison Wesley.
Luce, Mary Frances (1998), "Choosing to Avoid: Coping with Negatively Emotion-Laden
Consumer Decisions," Journal of Consumer Research, 24 (March), 409-433.
Luce, Mary Frances, James R. Bettman, and John Payne (2000), “Trade-off Difficulty:
Determinants and Consequences for Decision Making,” Monographs of the Journal of
Consumer Research Series, Chicago, IL: University of Chicago Press, forthcoming.
Luce, Mary Frances, James R. Bettman, and John W. Payne (1997), "Choice Processing in
Emotionally Difficult Decisions," Journal of Experimental Psychology: Learning, Memory,
and Cognition, 23 (March), 384-405.
Morwitz, Vicki G., Eric Johnson, and David Schmittlein (1993), “Does Measuring Intent Change
Behavior?,” Journal of Consumer Research, 20 (June), 46-61.
Presser, Stanley and Howard Schumann (1980), “The Measurement of a Middle Position in
35
Attitude Surveys,” Public Opinion Quarterly, 44 (1), 108-123.
Ross, Ivan (1969), “Handling the Neutral Vote in Product Testing,” Journal of Marketing
Research, 6, (May), 221-222.
Ryan, Michael J., Thomas Buzas, and Venkatram Ramaswamy (1995), “Making CSM a Power
Tool,” Marketing Research, 7 (Summer), 11-16.
Schneider, K.C. (1985), “Uninformed Response Rate in Survey research,” Journal of Business
Research, (April), 153-162.
Schumann, Howard and Stanley Presser (1981), Questions and Answers in Attitude Surveys,
New York: Academic Press.
Schwarz, Norbert and Gerd Bohner (2000), “Construction of Attitudes,” Blackwell Handbook of
Social Psychology: Intrapersonal Processes, A. Tesser and N. Schwarz, editors, Oxford, UK:
Blackwell.
Shafir, Eldar (1993), “Choosing Versus Rejecting: Why Some Options are Both Better and
Worse than Others,” Memory and Cognition, 21 (July), 546-556.
Si, Steven X. and John B. Cullen (1998), “Response Categories and Potential Cultural Bias:
Effects of an Explicit Middle Point in Cross-Cultural Surveys,” The International Journal of
Organizational Analysis, 6 (3), 218-230.
Simonson, Itamar (1989), "Choice Based on Reasons: The Case of Attraction and Compromise
Effects," Journal of Consumer Research, 16 (September), 158-174.
Simonson, Itamar (1992), "The Influence of Anticipating Regret and Responsibility on Purchase
Decisions," Journal of Consumer Research, 19 (June), 105-118.
Simonson, Itamar and Stephen M. Nowlis (2000), “The Role of Explanations and Need for
Uniqueness in Consumer Decision Making: Unconventional Choices Based on Reasons,”
36
Journal of Consumer Research, 27 (June), 49-68.
Simonson, Itamar and Amos Tversky (1992), "Choice in Context: Tradeoff Contrast and
Extremeness Aversion," Journal of Marketing Research, 29 (August), 281-295.
Slovic, Paul (1975), “Choice Between Equally Attractive Alternatives,” Journal of Experimental
Psychology: Human Perception and Performance, 1, 280-287.
Solomon, Michael (1999), Consumer Behavior, Upper Saddle river, NJ: Prentice Hall.
Thompson, M. M., P. M. Zanna, and D. W. Griffin (1995), “Let’s not be Indifferent About
(Attitudinal) Ambivalence. In R.E. Petty & J. A. Krosnick (Eds.), Attitude Strength:
Antecedents and Consequences (pp 361 386). Hillsdale, NJ: Erlbaum.
Viscusi, W. Kip, Wesley A. Magat, and Joel Huber (1987), “An Investigation of the Rationality
of Consumer Valuations of Multiple Health Risks,” Rand Journal of Economics, 18 (4), 465-
479.
... By not including a neutral point in a scale, the respondent is compelled to make a decision. The argument is that the qualitative results between the conventional five/seven points scales and the scale used here are unaffected since if the respondents are truly neutral, then they will randomly choose one or the other, so forcing them to choose should not bias the overall results (Garland, 1991;Chang, 1994, Kahn et al, 2003. It is also suggested that the exclusion of a neutral point will draw the respondent to make a decision one way or the other. ...
Article
Full-text available
Despite growing interest on the issues of communication, trust and commitment, studies examining the interplay between all of these three variables are lacking. This paper attempts to address this gap. It draws on survey data involving 244 employees from a medium-size food processing organization operating in NSW (Australia). The study explored relationships between communication, trust and commitment. Trust was measured by a six-items composite scale assessing overall beliefs in good intentions of organization participants as well as the degree of faith/trust in various actors in the organization, including co-workers and managers at various levels of the hierarchy. Correlation analysis revealed that perceived effectiveness of communication between management and employees, commitment & pride in working for the company and trust were significantly interrelated. However, the relationship between Trust and Communication was the strongest, with commitment also showing a significant relation to Trust. On the other hand, the relationship between commitment and communication was relatively weaker. The results demonstrate the importance of effective communication within organizations as it relates to trust and organizational commitment. In particular, the study shows that trust and commitment do not just happen; they are forged and maintained through effective communication. Implications for management practice and future research are discussed.
Article
Consumer choice is often influenced by the context, defined by the set of alternatives under consideration. Two hypotheses about the effect of context on choice are proposed. The first hypothesis, tradeoff contrast, states that the tendency to prefer an alternative is enhanced or hindered depending on whether the tradeoffs within the set under consideration are favorable or unfavorable to that option. The second hypothesis, extremeness aversion, states that the attractiveness of an option is enhanced if it is an intermediate option in the choice set and is diminished if it is an extreme option. These hypotheses can explain previous findings (e.g., attraction and compromise effects) and predict some new effects, demonstrated in a series of studies with consumer products as choice alternatives. Theoretical and practical implications of the findings are discussed.
Article
Choice conflicts between one's important values may cause negative emotion. This article extends the standard effort-accuracy approach to explaining task influences on decision processing by arguing that coping goals will interact with effort minimization and accuracy maximization goals for negatively emotion-laden decision tasks. These coping goals may involve both a desire to process in a thorough, accurate manner and a desire to avoid particularly distressing aspects of processing. On the basis of this extended framework, the authors hypothesized and found in 3 experiments that decision processing under increasing negative emotion both becomes more extensive and proceeds more by focusing on one attribute at a time. in particular, increased negative emotion leads to more attribute-based processing at the beginning of the decision process. The results are inconsistent with views that negative emotion acts only as an incentive or only as a source of decision complexity.
Article
Using a survey of 500 members of the general public and 500 lawyers from four cities, the authors examine uninformed response error in survey research. By asking respondents to evaluate the past performance of a fictitious public agency, the authors examine the impact of interest in the topic, the nature of the surrounding material, and the presence of a "don't know" option on the tendency to provide an uninformed response.
Article
A frequently encountered consideration in the construction of social survey opinion questions is whether respondents should be offered the choice of a middle response option that states a neutral response. This paper reports the results of six split-ballot tests in each of which a random half of the sample was offered the middle option and the other half was not. As in many earlier studies, the explicit offer of the middle option was found substantially to increase the proportion of respondents expressing a neutral response. An analysis of whether the size of this effect differed according to the socio-demographic subgroup of the respondent found few such interactions to be significant.
Article
This study reports an empirical test to determine which of two methods for allocating neutral responses (e.g., "no preference" or "no difference" for paired-comparison testing, or "just right" for a product at-tribute evaluation in a monadic test) better approximates the preference distribution that would have been ob-tained had respondents been forced to make a prefer-ence choice. The two methods compared are those apparently most often used for allocating neutral responses into prefer-ence responses: (a) equally divide the neutral responses and (b) divide the neutral responses into the preference categories in proportion to the obtained ratio of prefer-ences (hereafter called the proportionate method). Note, however, that there is at least one other method for allocating the neutral response [1] that involves a modi-fication of Thurstone's Law of Comparative Judgment [3]. According to the author, that approach produces results somewhere between the two more commonly used methods. Although the data to be discussed here were obtained in a monadic product-concept test situation, the findings recently reported by Odesky [2] on this problem relative to paired-comparison product testing are relevant. Odesky conducted a mail survey of homemakers and compared ten methods of asking paired-comparison preferences. Two of the ten methods were forced prefer-ence methods, but the remaining eight allowed various neutral responses. He concluded that "respondents who claim neutrality appear to exhibit the same preference patterns as those who express a preference;" therefore, he recommended either dividing no preference responses proportionate to preference responses or disregarding them altogether. He reached this conclusion because the proportionate method distributed the neutral re-sponses so that the derived distributions were not statis-tically different from the 'forced' preferences. He did not, however, compare distributions obtained by equally dividing the neutral responses to those obtained in the 'forced' preference situation. such a comparison using Odesky's data, the differences between the two distributions were not significant either (chi square <1). Therefore, Odesky's data provide no more support for the proportionate than for the divide-equally method. It remains unclear which of the two methods better approximates the forced preference dis-tribution.