Preferences for expressing preferences: People prefer finer evaluative distinctions for liked than disliked objects

��

Preferences for Expressing Preferences: People Prefer Finer EvaluativeDistinctions for Liked Than Disliked Objects

Rachel Smallman, Brittney Becker, Neal J. Roese

PII: S0022-1031(13)00209-6DOI: doi: 10.1016/j.jesp.2013.12.004Reference: YJESP 3135

To appear in: Journal of Experimental Social Psychology

Received date: 21 January 2013Revised date: 4 December 2013

Please cite this article as: Smallman, R., Becker, B. & Roese, N.J., Preferences for Ex-pressing Preferences: People Prefer Finer Evaluative Distinctions for Liked Than DislikedObjects, Journal of Experimental Social Psychology (2013), doi: 10.1016/j.jesp.2013.12.004

This is a PDF file of an unedited manuscript that has been accepted for publication.As a service to our customers we are providing this early version of the manuscript.The manuscript will undergo copyediting, typesetting, and review of the resulting proofbefore it is published in its final form. Please note that during the production processerrors may be discovered which could affect the content, and all legal disclaimers thatapply to the journal pertain.

http://dx.doi.org/10.1016/j.jesp.2013.12.004

http://dx.doi.org/10.1016/j.jesp.2013.12.004

ACC

EPTE

D M

ANU

SCR

IPT

ACCEPTED MANUSCRIPTPreferences for Expressing Preferences--1

Preferences for Expressing Preferences:

People Prefer Finer Evaluative Distinctions for Liked Than Disliked Objects

Rachel Smallman & Brittney Becker

Texas A&M University

Neal J. Roese

Northwestern University

Key Words: Preference, Categorization, Evaluation, Rating, Attitude

Correspondence:

Address correspondence to Rachel Smallman, Department of Psychology, Texas A&M

University, 4235 TAMU, College Station, TX 77843-4235; email: [email protected]

ACC

EPTE

D M

ANU

SCR

IPT


Abstract

Past research showed that people draw finer categorical distinctions for liked than disliked

objects, such that a wine lover, for example, sees greater detail and nuance among types of wine

than does a non-lover. In the present research, a similar pattern was found in evaluative

categorization (i.e., distinguishing between “somewhat liked” vs. “liked” vs. “greatly liked” etc.).

Across 5 experiments, respondents used finer evaluative distinctions (operationalized as more

versus fewer response options in a rating scale) when conveying attitudes about liked versus

disliked items. This effect extended to the level of mental representation and was moderated by

need for cognition, indicating the key role of elaboration (people elaborate more on liked vs.

disliked objects). These findings imply the potential usefulness of unbalanced rating scales (i.e.,

containing more scale points on the positive than negative side) so that respondents may better

express the nuances of their attitudes.

ACC

EPTE

D M

ANU

SCR

IPT


Preferences for Expressing Preferences:

People Prefer Finer Evaluative Distinctions for Liked Than Disliked Objects

Categorization is a basic component of cognition. From infants sorting blocks into piles

to biologists subdividing animals into phyla and species, categorization is an essential aspect of

human thought. However, with so many ways to divide things, how many categories are

typically used? Although many factors influence categorization, an important one involves

preference, such that liked objects demand finer distinctions than disliked objects (Smallman &

Roese, 2008, 2009). For example, when people who dislike science-fiction categorize television

shows, the broad category of “sci-fi” will suffice. To a science-fiction fanatic, however, Star

Wars, Stargate, and Star Trek are so utterly distinctive as to demand their own unique categories.

The current research extended this preference-categorization link into the evaluation domain, and

in particular focuses on the way in which people see shades of difference in their attitudes toward

various objects. For example, some people may find two evaluative categories sufficient to

express their attitude toward sci-fi shows (e.g., “somewhat favorable”; “extremely favorable”)

whereas others might demand additional evaluative categories (e.g., “barely favorable”;

“somewhat favorable”; “favorable”; “extremely favorable”). The results of five experiments

suggest that when assessing liked (vs. disliked) objects, people prefer a greater range of

distinctions among degrees of liking.

The ideas behind this research converge from traditions within social and cognitive

psychology. We begin with the supposition that people have a general tendency to think about,

ponder, and reflect upon that which they love. They spend time discovering and appreciating the

subtleties among objects related to their preference, as in the case of the wine lover drawn to

discover innumerable details about vineyards and vintages. In keeping with previous attitude

ACC

EPTE

D M

ANU

SCR

IPT


research, we use the term elaboration to denote information processing in which attitude-relevant

ideas are compared, connected, and synthesized (Petty & Cacioppo, 1986). Elaboration involves

relatively effortful consideration of detail and nuance. Elaboration can vary moment by moment,

but the tendency to engage in elaboration also varies across individuals, as captured by the need

for cognition scale (NFC; Cacioppo & Petty, 1982).

The role of elaboration in the link from preference to categorization may derive at least in

part from the Law of Effect (Thorndike, 1898), which states that rewarded behaviors tend to be

repeated. Contemporary interpretations of the Law of Effect position positive affect as the

rewarding feeling that evokes approach behavior (Carver, 2003; Gable & Harmon-Jones, 2008),

which in many cases involves repeated approach toward similar enjoyable objects. Accordingly,

people find pleasure in engaging in their preferences, which invites repeated interaction (Hoch &

Deighton, 1989), and repeated opportunity to learn new details about the preferred object. In

essence, people enjoy elaborating on activities, objects, and people that bring them pleasure.

Previous research demonstrated the preference-categorization effect via associative

conditioning (Smallman & Roese, 2008). By repeatedly pairing novel symbols with positive or

negative IAPS images (see Hofman, De Houwer, Perugini, Baeyens, & Crombez, 2010), new

preferences were created in the laboratory. Symbols included hieroglyphics and hobo symbols,

and thus were novel but also (initially) affectively neutral. Participants conditioned to like the

symbols subsequently divided them into more categories than participants conditioned to dislike

the symbols. Notably, participants mainly used evaluative terms to describe their categories. That

is, 87% of the time participants used valenced adjectives (e.g., “inspiring” or “ominous”) to label

their groupings. This observation suggests a new but unexplored aspect of the preference-

ACC

EPTE

D M

ANU

SCR

IPT


categorization effect: Might people prefer a larger arsenal of evaluative distinctions when

expressing their attitudes about liked versus disliked objects?

The structure and function of attitudes have been studied since the beginning of

psychological research (Bohner & Nickel, 2011; Eagly & Chaiken, 1993), with self-report rating

scales most typically used in assessing explicit attitudes. Earlier research examined the number

of response options within such scales primarily in terms of optimizing internal reliability

(Garner, 1960; Komorita & Graham, 1965; Weng, 2004): Too many or too few response options

decrease reliability, but 5 to 7 response options are generally ideal. Yet across decades of

research, an unquestioned assumption has been that bipolar attitude scales should be balanced,

i.e., an equal number of response options should be placed to the left and right of the neutral

middle option (Himmelfarb, 1993; Krosnick, Judd, & Wittenbrink, 2005). The current research

was prompted, however, by our noticing web-based attitude scales created by laypersons that

were unbalanced, and always unbalanced such that they favored a greater number of response

options on the positive side. For example, a typical four-point scale might include “poor,” “so-

so,” “good,” and “great” (i.e., 1 negative option, 1 neutral option, and 2 positive options). These

lay-created scales might perhaps capture a general tendency of people to use finer evaluative

distinctions to capture attitudinal variation among liked versus disliked objects. The present

research examined this possibility.

Elaboration can help explain a relation between preference and evaluative categorization.

Liked objects invite elaboration, for the simple reason that it is pleasurable to do so: baseball

fans love to talk baseball and fashion mavens love to talk fashion. For things cherished, people

relish the details, revel in nuance, and linger over memories, stimulating categorical

differentiation. Variation across individuals who are higher versus lower in NFC would provide

ACC

EPTE

D M

ANU

SCR

IPT


evidence for the role of elaboration. High NFC individuals seek out, acquire, and reflect back

upon information from their environment to a greater extent than low NFC individuals

(Cacioppo, Petty, Feinstein, & Jarvis, 1996). They welcome and are intrinsically motivated to

engage in cognitively effortful activities. In contrast, low NFC individuals are cognitive misers

(Taylor, 1981) who avoid engaging in effortful cognitive activity unless extrinsically motivated

(Amabile, Hill, Hennessey, & Tighe, 1994; Thompson, Chaiken, & Hazlewood, 1993).

Accordingly, we expected that high NFC individuals should be more likely to engage in

elaboration regardless of preference; in essence, their high intrinsic motivation for effortful

cognitive processing should weaken the preference-categorization effect. By contrast, low NFC

individuals should be less likely to engage in elaboration, but will be stimulated to elaborate

when there is high external motivation to do so, specifically, when thinking about preferred

objects. As a result, we expected that lower NFC individuals would be more likely to show the

pattern of using finer evaluative distinctions for liked versus disliked items.

Five experiments tested these ideas. Experiments 1a and 1b documented the basic effect

that preference influences how many evaluative distinctions participants felt were necessary to

convey their attitudes about liked versus disliked objects. Experiment 2 used a different

paradigm to confirm this basic pattern. Experiment 3 clarified the pattern further by showing that

the effect of preference on evaluative categorization is not merely due to a style of verbal

presentation, but rather extends to basic differences in mental representation. Finally, Experiment

4 revealed that NFC moderated the effect of preference on evaluative categorization, thus

providing evidence for the role of elaboration.

ACC

EPTE

D M

ANU

SCR

IPT


Experiments 1a and 1b

Experiments 1a and 1b examined how many evaluative distinctions participants picked to

communicate their opinions of liked versus disliked objects. Just as the number of “stars” are

sometimes used by critics to convey movie quality, participants were asked to become amateur

critics and decide how many “stars” were required to communicate meaningful distinctions to a

wider audience. Preference was manipulated on a within-subject basis; participants made

separate judgments for liked versus disliked objects. The dependent variable focused on how

many distinctions participants required to review each object adequately. In Experiment 1a,

participants created their own scales and provided category labels for each scale point. To bypass

the confounding role of vocabulary size or accessibility, Experiment 1b presented to participants

pre-constructed scales of varying lengths, from which they made a selection.

An alternative explanation is that the preference for finer distinctions could simply be a

function of whichever valence is more characteristic of the majority of objects in the category.

That is, when people believe that a category contains more liked objects (e.g., most food is good)

or disliked objects (e.g., most music is bad), they might demand more evaluative distinctions

simply to better accommodate the increased volume of valenced objects (e.g., more positive

distinctions for food and more negative distinctions for music). We tested this possibility by

having participants estimate the proportion of each category that contained liked (vs. disliked)

items.

Method

Undergraduate students (Experiment 1a, N = 35; Experiment 1b, N = 80) participated for

course credit. They judged how many categories were needed for 8 liked and 8 disliked objects

from the following domains: movies, clothing, music, food, concerts, university courses,

ACC

EPTE

D M

ANU

SCR

IPT


television shows, and sports. Instructions were: “On the next page, you’ll see a list of classes of

things (e.g., categories such as music, movies, clothes, etc). Your job is to assume that you are

going to be a CRITIC. If you were a critic (let’s say for movies), how many different scale points

would you need in order to communicate effectively to others (i.e., to make USEFUL

recommendations to other people)?”

The last sentence contained the preference manipulation. In the liked condition,

participants read: “Focus on things you like. For example, the music, clothes, and foods that you

enjoy.” In the disliked condition, participants read: “Focus on things you don’t like. For example,

the music, clothes, and food that you dislike.”

Experiments 1a versus 1b employed different measures of evaluative categorization.

Experiment 1a used an open-ended measure, in which participants created and labeled their own

evaluative scale (between 1 and 6 rating category labels) for each type of liked and disliked

object. Participants also estimated the proportion (percentage) of each category that consisted of

“liked” versus “disliked” items. In Experiment 1b, participants picked an ideal rating scale from

a list of pre-constructed scales. For each object, participants saw five unipolar rating scales,

ranging from 2 to 6 points. The point labels were drawn from the most common responses in

Experiment 1a (that were appropriate for all 8 objects). Each scale had identical endpoints

(neutral and worst for negative scales; neutral and best for positive scales). Scales were

lengthened by adding a new midpoint to the previous scale. See Appendix for scales. Participants

selected scales for liked versus disliked objects separately (as blocks), with block order

randomized between participants.

Results

ACC

EPTE

D M

ANU

SCR

IPT


In preliminary analyses, the order of presentation (liked vs. disliked) was not significant

as a main effect nor as a moderator and was omitted from the analyses presented below. In

Experiment 1a, we counted the number of rating points for which participants provided distinct

labels. Across 8 items, participants created scales that included more rating points for liked

versus disliked items (M = 3.3, SD = 0.75 vs. M = 2.9, SD = 0.72), F(1, 34) = 7.15, p = .011, d =

.90. Similarly, in Experiment 1b, participants selected unipolar rating scales with a greater

number of options when focusing on liked versus disliked items (M = 4.3, SD = 0.70 vs. M = 3.9,

SD = 0.82), F(1,79) = 33.2, p < .001, d = 1.29.

Further, in Experiment 1a, participants reported a greater proportion of liked items across

categories (55%; t(34) = 2.20, p = .035, d = .75). However, the proportion of liked items was not

significantly correlated with the desired number of rating points for liked items (r = .03, p = .88).

Similarly, the proportion of disliked items did not correlate with the desired number of rating

points for disliked items (r = .09, p = .62), thus ruling out the explanation that more evaluative

distinctions are demanded simply to better accommodate the increased volume of valenced

items1.

These findings provide initial evidence that the preference-categorization effect applies to

evaluative categorization, and the manner in which people express and communicate their

attitudes towards liked and disliked objects. Additionally, Experiment 1b eliminated vocabulary

accessibility as an alternative explanation. That is, preference may involve a richer vocabulary,

such that the individual may retrieve more relevant category labels from memory when

evaluating liked versus disliked objects. However, even when given pre-constructed scales

(rather than creating and labeling themselves), participants preferred more evaluative distinctions

for liked compared to disliked objects.

ACC

EPTE

D M

ANU

SCR

IPT


Experiment 2

The previous experiments showed that people prefer more evaluative categories for liked

versus disliked items. In essence, we have demonstrated a “preference” for how people express

their preferences. As a further demonstration of how initial preference shapes the way attitudes

are expressed, Experiment 2 assessed participants’ self-reported characterizations of the efficacy

of particular ways of expressing an attitude. That is, participants rated the usefulness of several

unipolar scales of differing lengths. In this way, we hoped to provide converging support for the

idea that people prefer more numerous gradations of attitude for liked than disliked objects.

Experiment 2 embraced a 2 x 2 between-subjects design that manipulated both preference (like

vs. dislike) and rating scale length (many versus fewer scale points). The dependent measure was

the degree to which participants found the rating scale to be useful in expressing their opinion.

Method

Participants (N = 104) were recruited using Amazon’s Mechanical Turk. In a between-

subjects design, participants focused on liked or disliked objects within a given category

(electronics, food, movies, or music). They then saw a rating scale with many (5) or few (3)

labeled scale points (three and five-point scales from Experiment 1b; see Appendix). Finally,

participants answered four questions regarding the scale’s efficacy: how useful the scale was,

how easy the scale would be to use, how well someone else could understand the participant’s

ratings, and how well the participant could understand someone else’s ratings (α = .90).

Results

A 2 (preference: liked versus disliked) x 2 (scale points: 5 vs. 3 points) ANOVA revealed

a significant interaction, F(1, 100) = 8.67, p = .004 (see Figure 1). When focusing on liked

objects, participants gave higher efficacy ratings to 5-point rather than 3-point scales (M = 4.8,

ACC

EPTE

D M

ANU

SCR

IPT


SD = 1.01 vs. M = 3.7, SD = 1.08), t(51) = 3.47, p = .001, d = .97. By contrast, when focusing on

disliked objects, the effect disappeared, such that participants gave similar efficacy ratings for

both scale versions (M = 4.2, SD = 1.25 vs. M =4.4, SD = 0.93), t(49) = 0.86, p = .394, d = .25.

Experiment 3

We have shown that when people express their attitudes, they demand greater ways of

differentiating among liked versus disliked objects. A wine lover demands more ways of

expressing variation in quality and enjoyment of wine than does a non-lover. A key question has

remained hidden to the methods used thus far, which is how deep the effect goes in terms of

mental representation. That is, the observed effect might be due a basic pattern of mentally

representing liked objects with finer evaluative distinctions. For example, the wine connoisseur

might be able to retrieve numerous particular wines at each of several unique levels of evaluation

(e.g., good vs. great vs. exceptional), whereas a non-lover might only be able to retrieve

particular wines at only one or a few levels of evaluation (e.g., bad). An alternative possibility is

that both lovers and non-lovers are equivalent in how they both represent and retrieve particular

instances at each evaluative level. Instead, they differ only in how they choose to present their

attitudes to others (i.e., in the labels and terms they use to express their attitudes verbally). The

previous experiments have only tapped this latter aspect of verbal expression. An alternative

approach, one that gets closer to the level of mental representation, is to have participants

consider a fixed range of evaluative categories and to consider how many items exist within each

evaluative category. By looking at variation in the distribution of the items, we can test whether

the effect of preference on evaluative categorization extends to mental representation.

ACC

EPTE

D M

ANU

SCR

IPT


In the present experiment, participants received a 6-point rating scale and determined the

percentage of items that fell into each level of that scale. Participants did this either for liked

versus disliked objects. If liked and disliked objects are mentally represented with similar

degrees of differentiation, then participants should be similar in their item distribution across the

scale points. If, however, liked objects involve greater differentiation at the level of mental

representation, the distributions should differ. Namely, liked items should be distributed more

evenly across the scale points, whereas disliked items should be distributed into a smaller cluster.

Method

Ninety-three students participated for course credit. In a between-subjects design,

participants were randomly assigned to either a liked or disliked cuisine (American, Italian,

Asian, or Mexican) and again played the role of food critic. The dependent measure involved a

6-point unipolar rating scale (six-point scales from Experiment 1b; see Appendix); participants

determined the percentage of food items within their selected cuisine that fell into each of the 6

scale points.

Results

A repeated-measures ANOVA with preference (liked vs. disliked) as the between-

subjects factor and scale point (6 scale labels) as the within-subject variable showed a non-

significant effect for preference (F(1, 91) = 0.00, p = 1.00, d = 0.0) and a significant effect for

scale point (F(1, 91) = 4.93, p = .029, d = .46). However, these main effects were qualified by a

significant Preference x Scale Point interaction (F(1, 91) = 93.10, p < .001)2.

As seen in Figure 2, the distribution of items across the scale points varies as a function

of preference. Relative to liked items, a significantly higher proportion of disliked items were

placed in the first (M = 45.7%, SD = 28.41% vs. M = 7.3%, SD = 5.60%; t(91) = 8.72, p < .001, d

ACC

EPTE

D M

ANU

SCR

IPT


= 1.83) and second (M = 16.3%, SD = 17.09% vs. M = 10.3%, SD = 6.64%; t(91) = 2.19, p = .03,

d = .46) scale points. In contrast, a significantly higher proportion of liked items were distributed

across the remaining four scale points; participants placed more liked items in the third (M =

15.4%, SD = 7.89% vs. M = 10.2%, SD = 7.84%; t(91) = 3.15, p = .002, d = .66), fourth (M =

21.0%, SD = 9.38% vs. M = 10.7%, SD = 10.39%; t(91) = 4.97, p < .001, d = 1.04), fifth (M =

19.4%, SD = 7.98% vs. M = 7.9%, SD = 7.10%; t(91) = 7.33, p < .001, d = 1.54), and sixth (M =

26.7%, SD = 20.21% vs. M = 9.1%, SD = 10.11%; t(91) = 5.43, p < .001, d = 1.14) scale points.

Accordingly, when considering liked food items, participants saw (and mentally

represented) finer distinctions among them. They distinguished between “Good”, “Really Good”,

“Great”, and “Best” items, and thus distributed a greater proportion of items across 4 of the 6

scale points. However, when focused on disliked food items, a much coarser distinction was

apparent; they may only distinguish between “Neutral” and “Poor” items, and therefore

distribute the majority of disliked items into these two categories. Interestingly, liked (vs.

disliked) items were distributed over scale points representing greater extremity (vs. neutrality).

In essence, liked items were more numerously positioned in the extremely liked range but

disliked items were less likely to appear in the extremely disliked range. This experiment thus

provided clarifying evidence that the effect of preference on evaluative categorization is not

merely due to a style of verbal presentation, but rather extends deeper to basic differences in

mental representation.

Experiment 4

The present experiment provides evidence for the role of elaboration in the effect of

preference on evaluative categorization by testing variation as a function of NFC. As noted

ACC

EPTE

D M

ANU

SCR

IPT


previously, high NFC individuals welcome and are intrinsically motivated towards effortful

cognitive activity whereas low NFC individuals avoid effortful cognitive activity unless they are

extrinsically motivated to do so (Amabile et al., 1994; Thompson et al., 1993). Accordingly, we

predicted that high NFC individuals should be more likely to engage in elaboration regardless of

preference (due to their high intrinsic motivation). By contrast, low NFC individuals may be

stimulated to elaborate only when there is high external motivation to do so (as when thinking

about preferred objects). Overall, we expected that low NFC individuals would be more likely

than high NFC individuals to show the observed preference-categorization effect. To test these

ideas, preference was manipulated as in the previous experiments alongside a self-report measure

of NFC.

An additional aspect of this experiment is the inclusion of a basic categorization

dependent measure (the volume of categorical distinctions created by participants). We included

this additional measure to build a bridge between earlier research on preference and basic

categorization (Smallman & Roese, 2008) and the current focus on evaluative categorization. As

will be seen, these constructs turned out to be independent.

Method

Participants (N = 144) were recruited using Amazon’s Mechanical Turk. In a between-

subjects design, participants were randomly assigned to either a liked or disliked cuisine

(American, Italian, Japanese, Mexican, Seafood, or Vegan). Participants then completed four

manipulation check questions about this cuisine (5-point likert scales with strongly disagree and

strongly agree as anchors; positive feelings, negative feelings, happiness, enjoyment; α = .96).

Next, participants completed two categorization measures. As in Experiment 1b,

participants played critic and picked their preferred rating scale from a selection of 6 unipolar

ACC

EPTE

D M

ANU

SCR

IPT


rating scales (ranging from 2 to 7 scale points; see Appendix). Scales were identical to previous

studies, with the exception of the 7-point scale (included to reduce a possible ceiling effect).

Participants then completed an open-ended categorization measure in which they imagined going

out to dinner at a restaurant serving their selected cuisine, and described what they would expect

to see on the menu. Following previous categorization research (Medin, Lynch, Coley, & Atran,

1997; Rosch, Mervis, Gray, Johnson, & Boyes-Braem, 1976), each discrete food item was scored

according to category level (superordinate, basic, subordinate, or uncodable). Category level

served as a guide to coders, orienting them to what to look for. We did not anticipate, nor did we

find, variation as a function of category level, so a tabulation across all levels resulted in a single

index of category volume. Two independent coders scored their responses; disagreement was

resolved through discussion.

Finally, participants completed the 18-item NFC scale (Cacioppo, Petty, & Kao, 1984).

Sample items include “I find satisfaction in deliberating hard and for long hours” and “I usually

end up deliberating about issues even when they do not affect me personally” (α = .94).

Results

The preference manipulation was successful, in that liked cuisine was rated more

favorably than disliked cuisine (M = 4.55, SD = 0.48 vs. M = 2.09, SD = 0.93), F(1, 142) =

408.6, p < .001, d = 3.37.

Replicating previous results, participants chose more scale points when thinking about

liked (M = 5.1, SD = 1.2) than disliked cuisine (M = 4.4, SD = 1.4), F(1, 142) = 12.68, p = .001,

d = 0.59. A similar pattern was found for the category volume index, such that participants used

more categories for liked (M = 5.4, SD = 4.3) than disliked cuisine (M = 3.3, SD = 3.8), F(1, 142)

ACC

EPTE

D M

ANU

SCR

IPT


= 10.10, p = .002, d = 0.53. Interestingly, these two dependent measures were uncorrelated (r =

.08, p = .37), and thus appear to represent distinct constructs.

To examine our prediction that preference would influence evaluative categorization to a

greater extent for low NFC individuals, we used regression procedures recommended by Cohen

and Cohen (1983). We entered preference (dummy coded; 0 = disliked cuisine, 1 = liked cuisine)

and NFC (mean-centered) in the first step, and the Preference x NFC interaction term in the

second step (Aiken & West, 1991). Regression coefficients are from the second step (see Figure

3).

For the model predicting the number of scale points selected by participants, there was a

significant main effect of preference (β = .728, t(140) = 3.47, p = .001), a non-significant main

effect of NFC (β = -.001, t(140) = .08, p = .94), and a significant Preference x NFC interaction (β

= -.032, t(140) = 2.08, p = .039). Simple slope analyses confirmed that preference was positively

related to the number of scale points selected for low NFC participants (β = 1.17, t(140) = 5.55, p

< .001) and unrelated to the number of scale points selected for high NFC participants (β = .28,

t(140) = 1.37, p = .17).

Similar results were obtained for the model predicting the number of categories included

in the open-ended categorization task. There was a significant main effect of preference (β =

2.14, t(140) = 3.20, p = .002), a non-significant main effect of NFC (β = .05, t(140) = 1.43, p =

.16), and a significant Preference x NFC interaction (β = -.103, t(140) = 2.13, p = .035). Simple

slope analyses confirmed that preference was positively related to the number of categories listed

for low NFC participants (β = 3.57, t(140) = 3.91, p < .001) and unrelated to the number of

categories listed for high NFC participants (β = .71, t(140) = .78, p = .44).

ACC

EPTE

D M

ANU

SCR

IPT


General Discussion

People see finer categorical differentiation for liked relative to disliked objects

(Smallman & Roese, 2008, 2009). The current research shows that this preference-categorization

effect also applies to evaluative categorization, i.e., rating scales used to communicate attitudes.

When expressing opinions about liked objects, people prefer scales with more points, thus

allowing them to make finer distinctions among liked objects. In contrast, when conveying

attitudes about disliked objects, people prefer scales with fewer points.

In the present research, participants created (Experiment 1a) and selected (Experiment

1b) scales with more evaluative distinctions for liked than disliked objects. Experiment 1b

included pre-labeled scales to rule out differences in vocabulary or accessibility. Beyond

showing a basic preference for how people express their preferences, Experiment 2 provided

converging evidence by showing that scales with more distinctions received higher efficacy

ratings when people considered liked (vs. disliked) objects. Participants considered these scales

to be more useful, easier to use, better able to express their own attitudes, and better for

interpreting someone else’s attitudes.

One alternative explanation for these findings is that a desire for finer distinctions might

be a function of whichever valence is more characteristic of the majority of objects in the target

category. When people believe there to be more liked than disliked objects, then they might

demand more evaluative distinctions to better accommodate the increased volume of valenced

objects. However, Experiment 1a ruled out this possibility in showing that regardless of whether

people liked the majority or minority of objects in a given class, they nevertheless desired more

options for describing the liked versus disliked objects.

ACC

EPTE

D M

ANU

SCR

IPT


Experiment 3 delved deeper into this effect, testing whether it results from differences in

either the mental representation of these distinctions or in how they are expressed. Experiment 3

showed that this effect extends to the level of mental representation. Specifically, when given 6

evaluative scale points, liked items were distributed more evenly across scale points, whereas

disliked items were distributed more succinctly. Although this points to the possibility that there

are differences in underlying representation, external factors may still influence this effect. For

example, if the situation is designed such that the scale would be used to make the items better

(i.e., activating an improvement motivation), it may warrant more negative evaluative

distinctions. Similarly, it the situation requires an individual to justify bad decisions (e.g., journal

editors or reviewers) or involves selecting an unavoidable negative outcome (e.g., a required

punishment), then finer degrees of “badness” would serve a purpose. Future research should

further explore these possibilities.

We also found evidence for the role of elaboration in the preference-categorization effect.

When engaging with desired items, people dwell on details, elaborate on nuances, and build on

memories, which stimulate categorical differentiation. We used variation across individuals in

need for cognition (Cacioppo & Petty, 1982; Cacioppo et al., 1996) to examine the role of

elaboration, and found that NFC moderated the preference-categorization effect. Whereas low

NFC individuals showed the predicted pattern for both traditional and evaluative categorization

measures, high NFC participants showed a much weaker pattern on these same measures. We

interpreted these results in line with previous evidence showing that low NFC individuals are not

intrinsically motivated to elaborate, and so would require extrinsic (i.e., situational) inducements

to elaborate (e.g., focusing on liked vs. disliked objects). By contrast, high NFC individuals are

intrinsically motivated to elaborate and hence engage in effortful cognitive activity on a chronic

ACC

EPTE

D M

ANU

SCR

IPT


basis. Accordingly, they should be less sensitive to extrinsic inducements to engage in

elaborative activity.

We began this research expecting that the impact of preference on evaluative

categorization would extend from earlier research on the impact of preference on basic

categorization (Smallman & Roese, 2008). Specifically, in Experiment 4 we included a basic

categorization measure, derived from earlier categorization research (Medin et al., 1997; Rosch

et al., 1976). Although both measures showed similar variation as a function of preference and

similar moderation by need for cognition, these categorization measures did not correlate. We

can conclude, then, that evaluative categorization is empirically distinct from basic

categorization, and future research might examine further differences between these constructs.

The current research speaks to how individuals represent and express their views

regarding liked and disliked objects. Another implication of this research is whether unbalanced

rating scales are better research tools compared to traditional balanced rating scales. Testing

whether people prefer unbalanced bipolar scales (with more points on the favorable side)

compared to balanced bipolar scales is an important next step for this research. A further step is

to examine the predictive utility of unbalanced scales. The current research only tested what

individuals prefer to use; this question centers on what works best, in terms predictive analytic

power.

The current findings are potentially important for attitudes theory. In particular, issues

regarding valence asymmetry. A widely recognized conclusion is bad is stronger than good. That

is, negative information weighs more heavily than positive information during impression

formation (Baumeister, Bratslavsky, Finkenauer, & Vohs, 2001; Cacioppo, Gardner, & Berntson,

1997; Rozin & Royzman, 2001). However, the current results suggest that people differentiate

ACC

EPTE

D M

ANU

SCR

IPT


positive information with greater nuance. A simplified version of our result might be that good is

more interesting than bad. The key difference between the present research and impression

formation research centers on the level of analysis of good versus bad. In our case, we focus on

pre-existing preferences, and chart variation across different good or bad objects within those

preferences. In contrast, impression formation involves meeting a new person and integrating

discrete bits of information varying in valence into a coherent summary (Anderson, 1965; Asch,

1946; Roese & Morris, 1999; Skowronski & Carlston, 1989). In this integration, negative

information weighs more heavily than positive. However, if a person becomes liked (i.e., the

impression is favorable), that person will be encountered more often (we seek out those we like)

and new nuances will be observed and understood via repeated interaction. Disliked persons are

avoided, nothing new is learned, and fewer categories are needed to differentiate among those

who are disliked (Fazio, Eiser, & Shook, 2004). Thus, bad is stronger than good at the level of

integrating information, but good is stronger than bad in inviting repeated elaboration once an

impression (or preference) is formed.

Overall, we have shown that people choose to express their attitudes with finer degrees of

evaluative distinction for liked versus disliked objects. A wine lover demands more ways to

express variation in wine than does a non-lover. This effect extends to a basic level of how

people think about the domain in question, be it food, sports, or science-fiction shows. That

people prefer to use scales with more positive versus negative options poses new questions for

both attitude research and companies that use ratings to gauge consumer demand.

ACC

EPTE

D M

ANU

SCR

IPT


References

Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and interpreting interactions.

Newbury Park, CA: Sage.

Amabile, T. M., Hill, K. G., Hennessey, B. A., & Tighe, E. M. (1994). The work preference

inventory: Assessing intrinsic and extrinsic motivational orientations. Journal of

Personality and Social Psychology. 66, 950-967.

Anderson, N. H. (1965), Averaging versus adding as a stimulus-combination rule in impression

formation. Journal of Experimental Psychology, 70, 394-400.

Asch, S. E. (1946). Forming impressions of personality. Journal of Abnormal and Social

Psychology, 41, 259 290.

Baumeister, R. F., Bratslavsky, E., Finkenauer, C., & Vohs, K. D. (2001). Bad is stronger than

good. Review of General Psychology, 5, 323–370.

Bohner, G., & Nickel, N. (2011). Attitudes and attitude change. Annual Review of Psychology,

62, 391-417.

Cacioppo, J. T., Gardner, W. L., & Berntson, G. G. (1997). Beyond bipolar conceptualizations

and measures: the case of attitudes and evaluative space. Personality and Social

Psychology Review, 1, 3–25.

Cacioppo, J. T., & Petty, R. E. (1982). The need for cognition. Journal of Personality and Social

Psychology, 42, 116-131.

Cacioppo, J. T., Petty, R. E., Feinstein, J. A., & Jarvis, B. G. (1996). Dispositional differences in

cognitive motivation: The life and times of individuals varying in need for cognition.

Psychological Bulletin, 119, 197–253

ACC

EPTE

D M

ANU

SCR

IPT


Cacioppo, J. T., Petty, R. E., & Kao, C. E (1984). The efficient assessment of need for cognition.

Journal of Personality Assessment, 48, 306-307.

Carver, C. S. (2003). Pleasure as a sign you can attend to something else: Placing positive

feelings within a general model of affect. Cognition and Emotion, 17, 241–261.

Cohen, J., & Cohen, P. (1983). Applied multiple regression/correlation analysis for the

behavioral sciences (2nd ed.). Hillsdale, NJ: Lawrence Erlbaum.

Eagly, A. H., & Chaiken, S. (1993). The psychology of attitudes. Fort Worth, TX: Harcourt

Brace.

Fazio, R. H., Eiser, J. R., & Shook, N. J. (2004). Attitude formation through exploration:

Valence asymmetries. Journal of Personality and Social Psychology, 87, 293–311.

Gable, P. A., & Harmon-Jones, E. (2008). Approach-motivated positive affect reduces breadth of

attention. Psychological Science, 19, 476-482.

Garner, W. R. (1960). Rating scales, discriminability, and information transmission.

Psychological Review, 67, 343-352.

Himmelfarb, S. (1993). The measurement of attitudes. In A. H. Eagly & S. Chaiken (Eds.), The

psychology of attitudes (pp. 23-87). Fort Worth, TX: Harcourt Brace.

Hoch, S. J., & Deighton, J. (1989). Managing what consumers learn from experience. Journal of

Marketing, 52, 1-20.

Hofmann, W., De Houwer, J., Perugini, M., Baeyens, F., & Crombez, G. (2010). Evaluative

conditioning in humans: A meta-analysis. Psychological Bulletin, 136, 390-421.

Komorita, S. S., & Graham, W. K. (1965). Number of scale points and the reliability of scales.

Educational and Psychological Measurement, 25, 987-995.

ACC

EPTE

D M

ANU

SCR

IPT


Krosnick, J. A., Judd, C. M., & Wittenbrink, B. (2005). The measurement of attitudes. In D.

Albarracin, B. T. Johnson, & M. P. Zanna (Eds.), The handbook of attitudes (pp. 21-76).

Mahwah, NJ: Erlbaum.

Medin, D. L., Lynch, E. B., Coley, J. D., & Atran, S. (1997). Categorization and reasoning

among tree experts: Do all roads lead to Rome? Cognitive Psychology, 32, 49-96.

Petty, R. E., & Cacioppo, J. T. (1986). The elaboration likelihood model of persuasion. In L.

Berkowitz (Ed.), Advances in experimental social psychology (Vol. 19, pp. 123-205 ).

New York: Academic Press.

Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and data

analysis methods. Thousand Oaks, CA: Sage.

Roese, N. J., & Morris, M. W. (1999). Impression valence constrains social explanations: The

case of discounting versus conjunction effects. Journal of Personality and Social

Psychology, 77, 437-448.

Rosch, E., Mervis, C. B., Gray, W. D., Johnson, D. M., & Boyes-Braem, P. (1976). Basic objects

in natural categories. Cognitive Psychology, 8, 382 -439.

Rozin, P., & Royzman, E. B. (2001). Negativity bias, negativity dominance, and contagion.

Personality and Social Psychology Review, 5, 296–320.

Skowronski, J. J., & Carlston, D. E. (1989). Negativity and extremity biases in impression

formation: A review of explanations. Psychological Bulletin, 105, 131-142.

Smallman, R., & Roese, N. J. (2008). Preference invites categorization. Psychological Science,

19, 1228-1232.

Smallman, R., & Roese, N. J. (2009). Valence-dependent self-scrutiny in judgments of event

impact. Social Cognition, 27, 834-846.

ACC

EPTE

D M

ANU

SCR

IPT


Taylor, S. E. (1981). The interface of cognitive and social psychology. In J. Harvey (Ed.),

Cognition, social behavior, and the environment (pp. 189-211). Hillsdale, N J: Erlbaum.

Thompson, E. P., Chaiken, S., & Hazlewood, J. D. (1993). Need for cognition and desire for

control as moderators of extrinsic reward effects: A Person × Situation approach to the

study of intrinsic motivation. Journal of Personality and Social Psychology, 64, 987-999.

Thorndike, E. L. (1898). Animal intelligence: An experimental study of the associative processes

in animals. Psychological Review Monograph Supplement, 2, 1-109.

Weng, L. (2004). Impact of the number of response categories and anchor labels on coefficient

alpha and test-retest reliability. Educational and Psychological Measurement, 64, 956-

972.

ACC

EPTE

D M

ANU

SCR

IPT


Footnotes

1. Multilevel modeling, using hierarchical linear modeling (Version 6.02; Raudenbush & Bryk,

2002), was used to examine the effects of preference and proportion on rating points after

controlling for individual differences in proportion liked on average. The results revealed a

significant, positive relation between preference and the number of rating points (b = .32, SE =

.12, p = .015). However, proportion was unrelated to the number of rating points (b = .002, SE =

.003, p = .44), as was the interaction between preference and proportion (b = -.001, SE = .004, p

= .72),

2. Because the 6 within-participant categories were linearly dependent (all cells summed to 100),

we also conducted a non-parametric test of this interaction effect. As expected, it was also

highly significant, χ2(5) = 47.1, p < .001.

ACC

EPTE

D M

ANU

SCR

IPT


Appendix

Rating scales and labels used in the current experiments. Experiment 1b included scales ranging

from two to six points, Experiment 2 used three and five-point scales, Experiment 3 used the six

point scales, and Experiment 4 included scales ranging from two to seven points.

Scales used to evaluate liked objects

Two-point scale: Neutral, Best

Three-point scale: Neutral, Good, Best

Four-point scale: Neutral, Good, Great, Best

Five-point scale: Neutral, Okay, Good, Great, Best

Six-point scale: Neutral, Okay, Good, Really Good, Great, Best

Seven-point scale: Neutral, Okay, Good, Really Good, Great, Excellent, Best

Scales used to evaluate disliked objects

Two-point scale: Neutral, Worst

Three-point scale: Neutral, Bad, Worst

Four-point scale: Neutral, Bad, Terrible, Worst

Five-point scale: Neutral, Poor, Bad, Terrible, Worst

Six-point scale: Neutral, Poor, Bad, Really Bad, Terrible, Worst

Seven-point scale: Neutral, Poor, Bad, Really Bad, Terrible, Horrendous, Worst

ACC

EPTE

D M

ANU

SCR

IPT


Figure 1. Scale efficacy ratings as a function of preference and number of scale points

(Experiment 2).

0

1

2

3

4

5

6

Like Dislike

Effi

cacy

Rat

ing

Preference

Five-Point Scale Three-Point Scale

ACC

EPTE

D M

ANU

SCR

IPT


Figure 2. Percentage of food items at each scale point as a function of preference (Experiment

3). For liked items, the scale points were: (1) Neutral, (2) Okay, (3) Good, (4) Really

Good, (5) Great, and (6) Best. For disliked items, the scale points were: (1) Neutral, (2)

Poor, (3) Bad, (4) Really Bad, (5) Terrible, and (6) Worst.

0

10

20

30

40

50

60

1 2 3 4 5 6

Perc

enta

ge o

f It

ems

Evaluative Scale Points

Dislike Like

ACC

EPTE

D M

ANU

SCR

IPT


Figure 3. Categorical differentiation (as measured by evaluative scale points and category

listing) as a function of preference and need for cognition (NFC) in Study 4.

1

2

3

4

5

6

7

Dislike Like

Eval

uat

ive

Scal

e Po

ints

Low NFC High NFC

1

2

3

4

5

6

7

Dislike Like

Cat

ego

ries

Lis

ted

Low NFC High NFC

ACC

EPTE

D M

ANU

SCR

IPT


Highlights

▪ Examines whether preference influences evaluative categorization (rating scales).

▪ Participants used more evaluative distinctions for liked vs. disliked objects.

▪ Finely differentiated scales received higher efficacy ratings only for liked objects.

▪ The effect was moderated by need for cognition, indicating the role of elaboration.

▪ These findings suggest the potential usefulness of unbalanced rating scales.

Documents

Preferences for expressing preferences: People prefer finer evaluative distinctions for liked than disliked objects