JOURNAL OF EXPERIMENTAL SOCIAL PSYCHOLOGY 25, 253-271 (1989)

Affect versus Evaluation in the Structure of Attitudes

STEVEN J. BRECKLER

Johns Hopkins University

AND

ELIZABETH C . WIGGINS
Barnard College of Columbia University

Received September 15, 1987

A distinction is made between affect and evaluation in the structure of attitudes.

Affect refers to emotional responses and feelings engendered by an attitude
object. Evaluation refers to thoughts, beliefs, and judgments about an attitude
object. In Study I, multiple measures of affect and evaluation were collected in
six attitude domains. Estimates of the disattenuated correlation between affect
and evaluation varied from .25 to .89, supporting discriminant validity of the
distinction. Affect and evaluation were both correlated with a global measure of
attitude, even when the effects of one were partialled from the other. Study 2
focused on the attitude domain of blood donation. The disattenuated correlation
between affect and evaluation was 52. Affect, but not evaluation, was correlated
with independent measures of mood. Self-reported behaviors relating to blood
donation were more strongly related to affect than to evaluation. However, the
relationship between affect and behavior diminished with increasing experience
in donating blood. These results have important implications for theories of
attitude structure, techniques of attitude measurement, and studies of attitude
change and the attitude-behavior relationship. 0 1989Academic press, IX.

The distinction between affective and cognitive components of attitu

is both a common one and an old one. It is rooted in the classical
This research was supported by BRSG Grant SO7 RR07041 awarded by the Biomedical
Research Support Grant Program, Division of Research Resources, National Institutes of
Health, by an American Red Cross Matching Fund Grant (Chesapeake Region), and by a
National Science Foundation Presidential Young Investigator Award (BNS-86-57093). Portions
of the data were reported at the 1986 Midwestern Psychological Association meetings. We
are grateful to Clinton DeSoto, Anthony Greenwald, Anthony Pratkanis, Hilly Rnbinsky,
and several anonymous reviewers for providing helpful comments on earlier drafts. Please
address correspondence and reprint requests to Steven .I. Breckler, Department of Psychology,
Johns Hopkins University, Baltimore, MD, 21218.

253
0022-1031BP $3.00
Copyright 0 1989 by Academic Press. Inc.
All rights of reproduction in any form reserved.
254 BRECKLER AND WIGGINS

dichotomy between feeling and knowing as two facets of human experience

(Hilgard, 1980; McGuire, 1985). It was used in many of the earliest
treatments of the attitude concept (Chein, 1951;” Kramer, 1949; Krech
& Crutchfield, 1948; Smith, 1947), and it has played a dominant role in
theories of attitude structure and change (Insko & Schopler, 1967; Katz
& Stotland, 1959; Rosenberg, Hovland, McGuire, Abelson, & Brehm,
1960).
The affective component of attitude refers to emotions and drives that
are engendered by a specific attitude object (Fleming, 1967; Izard, 1977).
In contrast, the cognitive component of attitude refers to the location
of an object of thought on one or more dimensions of judgment (McGuire,
1985). The term cognition in this context is misleading, because in its
broadest sense cognition can include emotion-related functioning (cf.
Breckler & Wiggins, 1989; Buck, 1985; Leventhal, 1984). Thus, we use
the term evaluation in place of cognition to emphasize that this component
of attitude refers to judgments about an attitude object.
Affect and evaluation can be the products of very different learning
experiences (Greenwald, 1968; Insko & Schopler, 1967; Triandis, 1971).
Evaluation is represented primarily in verbal or semantic form, whereas
affect is primarily associated with nonpropositional forms of representation
(Breckler & Wiggins, 1989; Greenwald, 1982; Zajonc, 1980; Zajonc &
Markus, 1984).
Despite strong theoretical precedence, only five studies have adopted
the explicit goal of validating the affect-evaluation distinction (Breckler,
1984; Kothandapani, 1971; Mann, 1959; Ostrom, 1969; Woodmansee &
Cook, 1967). Unfortunately, these studies have not provided strong and
consistent support for the distinction. Measures of affect and evaluation
have sometimes been so highly intercorrelated (e.g., Ostrom, 1969) that
McGuire (1985) described them as redundant evaluative constructs (i.e.,
lacking discriminant validity). At other times, the data did not indicate
convergent validity for measures of the two attitude components (e.g.,
Mann, 1959; Woodmansee & Cook, 1967). Breckler’s (1984) recent study
of attitudes toward snakes produced the strongest support, with an observed
correlation of .38 between affect and evaluation.
If affect and evaluation are truly distinct components of attitude, then
several important consequences for attitude theory and research can be
identified. First, different measurement scales, and perhaps very different
procedures, will be needed for the separate assessment of each attitude
component (cf. Breckler & Wiggins, 1989). Second, affect and evaluation
may be associated with unique attitudinal functions (Katz, 1960; Smith,
Brunei-, & White, 1956). For example, the primary function of affect
may be to rapidly mobilize and direct behavior (an adaptive function),
whereas evaluation may contribute primarily to greater efficiency in in-
formation processing tasks (a knowledge function). Third, affect and
 AFFECT VERSUS EVALUAT1Oi-d 255

evaluation may have unique influences on behavior (cf. Millar & Tesser,
1986; Wilson & Dunn, 1986). Fourth, affect and evaluation may be as-
sociated with distinct dynamic processes and mechanisms. For example,
classical conditioning procedures may be most effective in changing affect.
whereas persuasive appeals based on logical arguments may work better
in altering evaluations (cf. Batra & Ray, 1985).
Correlational data have typically been used to vahdate the affect-
evaluation distinction. Several studies have used the m~tjtrait-multimetho~
approach (Campbell & Fiske, 1959) to analyze correlations among measures
of affect and evaluation (Kothandapani, 1971; Ostrom, 2969). Other studies
have used exploratory or confirmatory factor analysis (Corsuch, 1983;
Long, 1983) to analyze the correlations (Breckler, 1984; Woodma~$e~
& Cook, 1967). Several investigators have recently reported confirmatory
factor analyses of the earlier multitrait-multimethod data (Bagozzi, 1978;
Breckler, 1983; Widaman, 1985).
Caution must be exercised when using correlational analysis to establish
discriminant validity of theoretical constructs. Random measurement error
and unreliability in the measured variables can attenuate an observed
correlation. Multiple measures and mathematical corrections can be used
to partly compensate for such effects (cf. AUen & Yen, 1979). An observed
correlation can also overestimate the population correlation when measures
of truly distinct constructs share common sources of measurement error..
These effects can partly be reduced by using dissimilar meas~r~rne~~
procedures and by directly removing or partiahing variance due to common
methods.
In addition to examining the zero-order correlation between affec
evaluation, evidence for discriminant validity would be provided
affect and evaluation show distinct relationships with other variables
(e.g., behavior): (b) measures of affect and evaluation imply a difference
in global attitude toward an object, (c) correlations between affect and
evaluation vary systematically across attitude domains, and (d) correlations
between affect and evaluation differ as a function of a third variable such
as past experience.
The present studies provide evidence bearing on the dist~nc~on
affect and evaluation in the structure of attitudes. These studies build
on previous research (Breckler, 1984) that provided initial support for
the distinction, but which left some fundamental issues unresolved. The
previous study was designed to make the affect-evahration correlation
as low as possible. This was done by using dissimilar attitude measures,
and by using an attitude object (a snake) that was expected to produce
a divergence between affect and evaluation. Although the previous in-
vestigation provided strong support for the affect-evaluation d~st~uct~~~~
use of a single attitude domain limits its generahzabihty .
A judicious sampling of several attitude domains was use
256 BRECKLER AND WIGGINS

present studies to determine the range and variability of the affect-

evaluation correlation. Multiple measures were used so that the correlations
could be corrected for attenuation due to unreliability in the measured
variables. In the first study, affect and evaluation in six attitude domains
were measured using parallel forms of semantic differential scales (Osgood,
Suci, & Tannenbaum, 1957). In the second study, equal-appearing interval
scales (Thurstone & Chave, 1929) were used to measure affect, evaluation,
and behavior relating to blood donation. Subjects in this study included
both donors and nondonors of blood. Thus, in Study 2 it was possible
to investigate differences in intercomponent correlations as a function
of direct experience with blood donation.
STUDY 1
The purpose of Study 1 was to test the generality of the affect-evaluation
distinction across attitude domains. Affect and evaluation were measured
in six attitude domains: blood donation, legalized abortion, computers,
nuclear weapons, standardized admissions tests, and college comprehensive
examinations. Additional measures of global attitude in each domain were
also taken. These topics were selected to represent the variety of attitude
domains typically investigated in attitude research.
Several criteria were used to test discriminant validity of the distinction
in each domain. First, the favorability implied by measures of affect and
evaluation were compared. Second, the correlation between affect and
evaluation was examined and corrected for attenuation due to unreliability.
Third, the correlations between each attitude component and global attitude
were computed. Fourth, partial correlations between each component
and global attitude (removing the effects of the other component) were
examined.
The extent of separation between affect and evaluation was expected
to vary across attitude domains. A very clear distinction would be supported
by (a) differences in favorability implied by affect and evaluation, (b)
corrected correlations substantially less than one, (c) differences in cor-
relations with global attitude, and (d) positive partial correlations with
global attitude. If fewer of these criteria are met, the distinction between
affect and evaluation would be less evident.
Method
Procedure
Measures of affect, evaluation, and global attitude for each of the six attitude domains
were contained in a booklet completed by 152 college students as part of a classroom
exercise. Subjects were informed that their responses would remain anonymous.
For each attitude domain, subjects completed two pages of 7-point scales drawn from
the evaluative dimension of the semantic differential (Osgood et al., 1957). On one page,
subjects were instructed to evaluate attributes of the attitude object (evaluation), and on
the other page they were instructed to rate how the attitude object made them feel (affect).
 AFFECT VERSUS EVALUATION 2.57

Ordering of the two pages was counterbalanced across subjects, and the two pages for a
given attitude domain never appeared consecutively. The semantic differential anchors
used with each attitude domain (Table I) were selected to be meaningful as applied to that
domain and to make sense in describing both evaluation and affect. A separate section of
the booklet contained single-item scales to measure global attitudes for each attitude
domain.

Measured Variables
Evaluation. Instructions at the top of each evaluation page requested the subject to think
about the attitude domain and to use the scales on that page to “indicate your beliefs”
about it. Subjects then rated their evaluation of the attitude object by using the 7-point
semantic differential scales to complete the stem, (Attitude object) is . . The final measure
of evaluation was calculated by summing responses to the semantic differential scales used
with the evaluation stem. (The appropriate scales were reversed so that larger numbers
always indicated more favorable evaluations.)
Affect. Instructions at the top of each affect page requested the subject to use the scales
on that page to indicate how the attitude object “makes you feel.” Subjects then rated
affect engendered by the attitude object by using the same 7-point semantic differential
scales to complete the stem, (Attitude object) makes me feel. . The final measure of
affect was calculated by summing responses to the semantic differential scales used with
the affect stem.
Global attitude. Subjects rated their global attitudes by completing a 7-point, single-
item. self-rating scale for each domain. Subjects were instructed to rate their attitudes
toward each of the topics listed on the page, using a response scale that ranged from - 3

TABLE I
SEMANTIC DIFFERENTIALANCHOM: STUDY 1

Attitude domain Anchors

Blood donation Bad /Good

Wise/Fodish
Important/Unimportant
SelfishjUnselfish
Safe/Unsafe
Legalized abortion Bad/Good
Wise/Foolish
Kind/Cruel
Selfish/Unselfish
Computers Bad/Good
Wise/Foolish
Useless/Useful
Impo~a~t/Un~mpo~an~
Nuclear weapons Bad/Good
Wise/Foolish
UseIess/IJsefuI
Safe/Unsafe
Standardized Bad/Good
admissions tests Wise/Foolish
College comprehensive Bad/Good
examinations Wise/Foolish
258 BRECKLER AND WIGGINS

(dislike very much) to +3 (like very much). Intermediate categories were labeled with
numbers, and the middle category was labeled “neutral.”

Results’
Results did not differ as a function of the order in which the affect
and evaluation measures appeared in the booklet. Descriptive statistics
for Study 1 are summarized in Table 2. Inspection of Table 2 reveals
that affect was less favorable than evaluations in the attitude domains
of blood donation, computers, and comprehensive examinations. The
zero-order correlations between affect and evaluation, also shown in
Table 2, varied substantially across attitude domains. The lowest correlation
was for computers (Y = .20), and the highest was for legalized abortion
(Y = .77). The average aEect-evaluation correlation across the six attitude
domains was .56.2 All six correlations were more than two standard
errors below 1.0.
Table 3 gives internal consistency reliability (coefficient (x) estimates
for the summed measures of affect and evaluation. Reliabilities were all
high, ranging from .69 to .88. These reliability coefficients were used to
estimate the disattenuated correlation between affect and evaluation. The
corrected affect-evaluation correlations (also shown in Table 3) were
calculated by dividing the zero-order correlation by the square root of
the product of the two component reliabilities (see Allen & Yen, 1979).
The corrected correlations ranged from .25 (computers) to .89 (legalized
abortion and nuclear weapons). The average corrected affect-evaluation
correlation across the six attitude domains was .72.
Table 4 shows correlations of the global attitude measure (single-item,
self-rating scale) with the measures of affect and evaluation. Both affect
and evaluation were strongly correlated with the global measure of attitude
in each domain. The affect-attitude correlations ranged from .48 (com-
prehensive examinations) to .76 (nuclear weapons), with an average of
.62. The evaluation-attitude correlations ranged from .45 (computers) to
.88 (legalized abortion), with an average of .73. The correlation between
global attitude and evaluation was higher than the correlation between
global attitude and affect for legalized abortion (t(147) = 7.53), nuclear
weapons (t(149) = 2.62), standardized admissions tests (t(146) = 3.35),
and college comprehensive examinations (t(148) = 3.01). It should be
noted that the variances of the affect and evaluation measures differed
for legalized abortion (t(148) = 4.5) and comprehensive examinations
(t(149) = 3.6). Thus, the smaller correlations between affect and attitude
may be attributed, in part, to the smaller variances of the affect measures.

’ Because of the large number of inferential statistics reported in this paper, a conservative
Type I error rate of .Ol was adopted throughout.
’ Correlations were tist converted to Fisher’s z, averaged, and then converted back
to I-.
AFFECT VERSUS EVALUATION 259
260 BRECKLER AND WIGGINS

TABLE 3
RELIABILITY AND CORRECTED CORRELATIONS FOR AFFECT AND EVALUATION MEASURES:
STUDY 1

Coefficient (Y

Attitude domain Affect Evaluation Corrected r

Blood donation .81 .79 .77

Legalized abortion .83 .88 .89
Computers .88 .7s .25
Nuclear weapons .80 .81 .89
Admissions tests .73 .85 .65
Comprehensive exams .69 .86 .54

However, this potential artifact does not account for the smaller affect-
attitude correlation for nuclear weapons and standardized admissions
tests, for which the affect and evaluation variances did not differ.
The zero-order correlations of global attitude with affect and evaluation
support convergent validity of the two attitude components; in every
attitude domain, affect and evaluation were both correlated with global
attitude. To further investigate discriminant validity of the affect-evaluation
distinction, partial correlations of global attitude with affect and evaluation
were calculated. The third column of Table 4 shows partial correlations
between global attitude and affect, partialling the effects of evaluation;
the fourth column of Table 4 shows partial correlations between global
attitude and evaluation, partialling the effects of affect. The partial cor-
relations are high and reliably different than zero, with the single exception
of the affect-attitude partial correlation in the domain of legalized abortion.
The partial correlation between global attitude and evaluation was mgher

TABLE 4
CORRELATIONS OF GLOBAL ATTITUDE WITH AFFECT AND EVALUATION: STUDY 1

Zero-order correlation Partial correlation

between global attitude between global attitude
and and

Attitude domain Affect Evaluation Affect Evaluation

Blood donation .64 .63 .41 .39
Legalized abortion .68 .88 .03* .76
Computers .55 .45 -52 .42
Nuclear weapons .76 .84 .41 .66
Admissions tests .58 .75 .34 .65
Comprehensive exams .48 .67 .31 .59

* ns, not significant.

 AFFECT VERSUS EVALUATION 261

than that between global attitude and affect in the domains of legalized
abortion (t(145) = 22.4), nuclear weapons (t(147) = 5.33), standardized
admissions tests (t(144) = 4,95), and comprehensive examinations @(146)
= 3.87). These results indicate that affect and evaluation each share
nonredundant variance with global attitude.
Discussion
The extent of separation between affect and evaluation varied sub-
stantially across the six attitude domains. Perhaps the clearest distinction
was associated with attitudes toward college comprehensive examinations.
In this domain, affect and evaluation differed in favorability, the corrected
correlation was substantially less than one, correlations with global attitude
differed, and both partial correlations with global attitude were reliable
and positive. Support for the a.Eect-evaluation distinction was also relatively
strong for attitudes toward blood donation, computers, and standardized
admissions tests.
In contrast, the attitude domain of legalized abortion provided relatively
weak support for discriminant validity because few of the criteria outline
above were satisfied (correlations of affect and evaluation with global
attitude differed, and only one partia1 correlation was reliable). Attitudes
toward nuclear weapons similarly provided equivocal support for the
affect-evaluation distinction.
These results suggest a set of dimensions along which correspondence
between affect and evaluation may vary. The greatest consistency between
affect and evaluation was found for legalized abortion and nuclear we
ons-attitude domains that are topical, controversial, and likely to
the object of frequent thought by college students. In contrast, the domain
in which consistency was lowest-comprehensive examinations-is one
that is much less likely to be the object of frequent thought.
It should be noted that the majority of results are based on correlationa
analysis. Many theoretically irrelevant factors can influence the magnitude
of an observed correlation, including reliability, variance, skewness, and
measurement error. Although efforts were made to control for many of
these factors, some caution must be exercised in the interpretation of
results.
STUDY 2
Study 2 focused on attitudes toward blood donation. Three equal-
appearing interval scales (Thurstone & Chave, 1929) were developed to
measure affect, evaluation, and behaviors relating to blood donation.
Because the semantic differential scales used to measure affect and eval-
uation in Study 1 shared substantial method variance, they were likely
to have inflated the affect-evaluation correlation. The equal-appearing
262 BRECKLER AND WIGGINS

interval scales developed for Study 2 had very little redundancy in content
and were therefore expected to provide a more accurate estimate of the
affect-evaluation correlation.
Another addition in Study 2 was the use of independently developed
measures of emotion-related functioning. The affective component of
attitude is conceptualized as representing emotional experience associated
with an attitude object (see Breckler & Wiggins, 1989). Separate assessment
of emotional responses to blood donation were therefore collected to
assess convergent validity of the affect scales.
The attitude domain of blood donation was selected for closer scrutiny
for two reasons. First, attitudes toward blood donation showed relatively
clear separation of atfect and evaluation in Study 1. Second, blood donation
represents a behavioral domain in which people vary widely in their
direct behavioral experience. Thus, in addition to investigating the affect-
evaluation correlation, Study 2 permitted an examination of the affect-
behavior and evaluation-behavior correlations. Discriminant validity of
the affect-evaluation distinction will gain additional support if the two
attitude components show distinct relationships with behavior.
Recent models of the attitude-behavior relationship assign a central
role to direct experience in the attitude-to-behavior process (Fazio, 1986).
In a number of studies, Fazio and his colleagues have demonstrated
greater correspondence between self-reported attitudes and action when
those attitudes have been formed on the basis of direct rather than
indirect experience (see Fazio & Zanna, 1981). Direct experience appears
to increase the accessibility of attitudes from memory (see Fazio, 1986).
If it is assumed that attitudes must be accessed from memory before
they can have an influence on behavior, it then follows that direct experience
should produce a stronger correspondence between attitudes and action.
The mental and neural representations of affect versus evaluation suggest
important differences in the determinants of their accessibility from memory
(Breckler & Wiggins, 1989). Thus, when beliefs and other semantic rep-
resentations associated with an attitude object are made especially salient,
behaviors may be more strongly influenced by evaluations than by affect.
However, affect may govern behavior to a greater extent when emotional
responses predominate.
Blood donation is a behavioral domain in which affective responses
play a dominant role. Indeed, nondonors commonly mention fear and
anxiety as among the primary reasons for not donating blood (Oswalt,
1977). In contrast, donors may repeat their behavior because of positive
emotional experiences (Piliavin, Callero, & Evans, 1982). However, blood
donation also becomes habitual and automatic after repeated expression
(cf. Piliavin, Evans, & Callero, 1982). Thus, direct experience with blood
donation may actually decrease the chronic accessibility of attitudes from
 AFFECT VERSUS EVALUATION 263

memory. As one gains experience in donating blood, it may become less

likely that attitudes will be accessed from memory and therefore less
likely that they will influence or guide behavior (cf. Ronis, Yates,
Kirscht, 1989).
Method
Subject Samples
Four samples of subjects completed questionnaire booklets containing measures of affect,
evaluation, behavior, and general attitude toward blood donation.
Sample 1 included 97 undergraduate, graduate (master’s level), and continuing education
students attending summer school at a large state university. Subjects participated (for
pay) as part of an unrelated experiment.
Sample 2 included 86 members of a university community (faculty, students, and staff)
who were participating in a campus blood drive. Subjects completed the questionnaire just
after donating blood.
Sample 3 incladed 141 faculty and staff members of the same university community,
who responded to a mailed campus survey regarding attitudes toward blood donation.
Sample 4 included 160 undergraduate students attending a large state university. Subjects
participated to satisfy a course requirement.

Measured Variables
Equal-Appearing Interval (EAZ) scales. Three scales were developed using the method
of equal-appearing intervals (Edwards, 1957; Thurstone & Chave, 1929). The EM-A&W
scale included 12 statements such as “Donating blood makes me feel uncomfortable,”
“Donating blood makes me feel indiierent,” and “Donating blood makes me feel generous.”
The EAZ-Evaluation scale included 14 statements such as “It is wrong to donate blood,”
“Donating blood saves lives,” and “Blood donation is an important civic duty.” The EM-
Behavior scale included 14 statements such as “I refuse to donate blood,” “I won’t go
out of my w-ay to donate blood,” and “I donate blood as often as possible.” Each scale
appeared on a separate page in the questionnaire booklet, the ordering of which was
counterbalanced across subjects. For each set of statements, subjects were instructed to
check those items with which they agreed. A score for each scale was calculated as the
median of equal-appearing interval scale values for checked items (scale values ranged
from 1 to 7).
Dgferential Emotions Scale (DES). Subjects in Samples 2 and 3 responded to a subset
of items from the Differential Emotions Scale (Izard, 1972). The reduced DES scale contained
34 randomly ordered adjectives. Subjects were instructed to rate the extent to which each
adjective described the way blood donation made them feel. Ratings were made on a 5
point scale (from very slightly or not at all to very strongly). These ratings were used to
form five subscales: Joy (joyful, enthusiastic, delighted, happy), Interest (attentive, con-
centrating, alert, engaged in thought), Surprise (surprised, amazed, startled, astorzished),
Distress (downhearted, lonely, upset, distressed), and Fear (scared, fear&d, frightened,
anxious).
Semantic Differential (SD) scales. Subjects in Sample 4 also completed both sets of
semantic differential scales described in Study 1. In the evaluation set (SD-Evaluation)
subjects responded on 7-point scales to the stem, Blood donation is (bad/good, wise/foolish.
important/unimportant, seffish/unselflsh, safe/unsafe). In the affect set (SD-Affect) subjects
responded with the same 7-point scales to the stem, Blood donation makes me feel.
Responses within each set were summed (after reversing items 2, 3, and 5).
264 BRECKLER AND WIGGINS

Additional questions. Subjects in all samples answered general demographic questions

(age, sex, education), a question asking whether the respondent had ever donated blood
(and if so, how many times), a self-rating scale of experience with blood donation, and
the same 7-point global attitude scale used in Study 1.

Results
Summary statistics for the equal-appearing interval, semantic differential,
mood, and global attitude measures are given in Table 5. As in Study
1, affect engendered by blood donation was less favorable than were
evaluations of blood donation. This effect was observed for the equal-
appearing interval measures (t(441) = 7.11) and for the semantic differential
measures (t(157) = 12.71). The variance for EAI-Affect was greater than
the variance for EAI-Evaluation (t(428) = 18.2). However, the variances
of the two semantic differential measures did not differ. Experience with
blood donation varied substantially across subjects. Number of prior
donations ranged from 0 (41.5% of the subjects had never donated blood)
to 50. Over 25% of the subjects had donated blood four or more times.
Affect versus Evaluation
The zero-order correlation between equal-appearing interval measures
of affect and evaluation was .24 (N = 430). Reliability estimates were
not available for these measures. However, a lower-bound estimate for
reliability was taken as the correlation of each scale with its corresponding
semantic differential measure (Sample 4). The correlation between EAI-
Affect and SD-Affect was .52: the EAI-Evaluation and SD-Evaluation

TABLE 5
SUMMARY STATISTICS FOR MEASURES OF AFFECT, EVALUATION, BEHAVIOR, AND MOOD:
STUDY 2

Scale N Mean SD Coefficient 01

EAI-Affect 430 4.78 1.38 -

EAI-Evaluation 430 5.24 0.68 -
EAI-Behavior 430 4.60 1.49 -
SD-Affect” 158 25.28 4.90 0.77
SD-Evaluation” 158 29.39 4.75 0.80
General attitude 430 4.91 1.86 -
Prior donations 480 4.00 7.36 -
Joy* 189 8.51 4.18 0.91
Interest” 189 7.98 3.83 0.83
Surpriseb 189 4.51 1.52 0.75
Distress* 189 4.88 2.34 0.80
Fea? 189 6.67 4.17 0.93

Note.N = sample size (listwise deletion of missing values); SD = standard deviation.

’ Sample 4.
b Samples 2 and 3.
 AFFECT VERSUS EVALUATION 265

correlation was .41. These correlations were used to correct the EAI
affect-evaluation correlation for attenuation due to unreliability. The
corrected correlation was S2. It should be noted that this correlation
represents an upper-bound estimate, because the reliability coefficients
used to calculate it were lower-bound estimates.
The correlation between semantic differential measures of affect and
evaluation was .65 (N = 158). The corrected correlation (see Table 5
for reliability estimates) was .83. These correlations are similar to the
ones observed for blood donation in Study 1.
The zero-order correlations of global attitude with affect and evaluation
are given in Table 6. These correlations suggest that global at~it~~e~
toward blood donation are more strongly determined by affect than by
evaluations. The attitude-affect correlation was higher than the attitude-
evaluation correlation for the equal-appearing interval measures (t(427)
= 5.77) and for the semantic differential measures (t(l55) = 3.76).
ever, the difference in correlations for the equal-appearing interval measures
may be attributed, in part, to differences in their variances.
The partial correlations between global attitude and the equal-appearing
interval measures of affect and evaluation were both reliably different
than zero. The partial correlation between global attitude and the semantic
differelitial measure of affect was also a reliable one. However, the partial
correlation between global attitude and the semantic differential measure
of evaluation (partialling affect) was not reliably different than zero. The
partial correlation between global attitude and affect was greater than
that between global attitude and evaluation for the equal-appearing ~te~a~
measures (t(425) = 6.78) and for the semantic differential measures (t(l53)
= 9.39).
Correlations between the attitude measures and the mood subscales
are given in Table 7. Measures of mood were correlated with the equal-

TABLE 6
CORRELATIONS OF GLOBAL ATTITUDE WITH MEASURES OF AFFECT AND EVALUATION:
STUDY 2

Zero-order correlation Partial correlation

between global attitude between global attitude
and and

Scale Affect Evaluation Affect Evaluation

Equal-appearing .Sl” .29* .54* .20*

interval (N =
430)
Semantic differential .51* .36* .47* - .02
(N = 158)

* II < .Ol.
266 BRECKLER AND WIGGINS

TABLE 7
CORRELATIONS BETWEEN ATTITUDE MEASURES AND MOOD MEASURES:
STUDY 2 (SAMPLES 2 AND 3)

Equal-appearing interval
Global
Mood subscale Affect Evaluation Behavior attitude

JOY .36" .13 .44* .50*

Interest .I5 .Ol .11* -.14
Surprise -.16 .I1 -.I4 -.08
Distress -.45* -.06 -.44* -.48*
Fear - .55* - .05 -.41* - ho*

Note. N = 176.
* p < .Ol.

appearing interval measure of affect but not with that of evaluation. The
EAI-Affect scale was positively correlated with Joy and negatively cor-
related with Distress and Fear. The same pattern of correlations was
observed between the mood subscales and the global measure of attitude.
Behavior and Prior Experience
The equal-appearing interval measure of behavior was strongly correlated
with number of prior donations (r = .46) and with self-ratings of general
experience with blood donation (r = 56). The EAI-Behavior scale was
more strongly correlated with EAI-Affect (r = .62) than with EAI-Eval-
uation (Y = .28; t(427) = 7.45). The EAI-Behavior scale was also strongly
correlated with the global measure of attitude (r = S7); the attitude-
behavior correlation remained high even when EAI-Affect (pr = -34)
and EAI-Evaluation (pr = 53) were partialled. The EAI-Behavior scale
was also positively correlated with the Joy and Interest mood subscales
and negatively correlated with the Distress and Fear mood subscales
(see Table 7). These results indicate a strong relationship between affect
and behaviors relating to blood donation.
Subjects were divided into three groups based on the number of prior
blood donations: Nondonors (zero donations, N = US), Novices (one
to three donations, N = 133), and Veterans (four or more donations, N
= 136). Summary statistics for the three groups are given in Table 8.
A two-way mixed-model analysis of variance indicated a reliable interaction
between prior experience and attitude component (affect vs evaluation)
(F(2, 436) = 51.68). Simple main-effects analyses and Duncan multiple-
range tests were used to further analyze the interaktion. The three groups
all differed in their affect toward blood donation (F(2, 436) = 116.5).
Nondonors had less favorable evaluations than did novices and veterans,
but the latter two groups did not differ in their evaluations (F(2, 436) =
3.13). Within the experience groupings, nondonors had more favorable
 AFFECT VERSUS EVALUATION 267

TABLE 8
AFFECT AND EVALUATION SUMMARY STATISTICS SUBJECTS GROUPED BY PAST EXPERIENCE:
STUDY 2

EAI-Affect EAI-Evaluation

Experience group Mean SD Mean SD

Nondonors 3.83 1.43 5.06 0.81

Novices 5.10 1.19 5.34 0.55
Veterans 5.54 0.78 5.35 0.40

Note. SD = standard deviation.

evahtations than affect (t(164) = 10.40), whereas veterans had more

favorable affect than evaluations (t(l37) = 2.72); novices did not differ.
Correlations among the measured variables did not re~ably diier across
the three groups. The correlation between EAI-Affect and EM-Evaluation
was not significant for Nondonors (Y = .1.5) and Novices (Y = .19), but
it was significant for Veterans (v = .20). Similarly, the partial correlation
between EAI-Evaluation and global attitude (partialling EAI-Affect) was
not significant for Nondonors (pr = .18) and Novices @r = .14), but it
was significant for Veterans (pv = .31). The partial correlation between
EAI-Affect and global attitude (partialling EAI-Evaluation) was s
for Nondonors (pr = .38), for Novices (pr = .51), and for Vet
= .44).
The relationship between attitudes and behavior as a function of prior
experience was investigated using hierarchical multiple linear regression
analyses (cf. Cohen & Cohen, 1983). The criterion variable was EAT-
Behavior, and the predictor variables were global attitude and number
of prior donations. Both predictor variables had strong linear relationships
with behavior, such that higher EAI-Behavior scores were associated
with more favorable global attitudes (partial .35, t(460) = 11.41)
and a greater number of prior donations (partia = 46, t(460) = 8.16).
However, these main effects were qualified a reliable interaction
between global attitudes and number of prior donations (partial
-0.04, t(459) = - 5.49). The negative partial regression coefficient for
the interaction indicates that the linear relationship between global ~t~it~~~
and EAI-Behavior decreases with increasing number of pi-& donations;
the linear relationship approaches zero as the number of prior donations
approaches 12.
Similar multiple regression analyses were conducted with the EAI-
Affect and -Evaluation measures and number of prior donations as pae-
dictors of EAI-Behavior. All three predictors were associated with ~osit~.~e
partial regression coefficients, indicating that higher EAH-Behavior scores
were associated with more favorable affect (partial B = .54), more favorable
268 BRECKLER AND WIGGINS

evaluations (partial B = .30), and greater number of prior donations

(partial B = .05). The only reliable interaction was between EAI-Affect
and number of prior donations (partial B = - .04). This result is similar
to that found with global attitude, such that the linear relationship between
EAI-Affect and EAI-Behavior decreases with increasing number of prior
donations; the linear relationship approaches zero as the number of prior
donations approaches 15.
Discussion
Results from Study 2 provided further support for the affect-evaluation
distinction. As in Study 1, affect toward blood donation was less favorable
than were evaluations. The equal-appearing interval scales of affect and
evaluation produced a correlation that was substantially less than one
and much lower than the correlation associated with semantic differential
measures. The lowered redundancy in content of the equal-appearing
interval scales may largely account for this difference in results.
Partial correlation analyses also supported the affect-evaluation dis-
tinction. Both affect and evaluation were correlated with global attitude,
even when the effects of one were partialled from the other. The pattern
of correlations, however, suggested that global attitudes toward blood
donation may be more strongly dominated by affect than by evaluation.
Dividing subjects on the basis of their prior experience revealed that
nondonors’ evaluations were more favorable than their affect toward
blood donation. However, veteran blood donors (four or more donations)
had more favorable affect than evaluations. These results are consistent
with studies of blood donor motivation that identify emotional responses
as strong determinants of blood donating behavior (Oswalt, 1977; Piliavin
et al., 1982).
Multiple regression analyses of the attitude-behavior relationship showed
that both affect and evaluation accounted for variance in self-reported
behaviors, However, the interaction between affect and number of prior
donations placed an important qualification on the interpretation of these
main effects-the linear relationship between affect and behavior decreased
with increasing experience. These results suggest that the affective com-
ponent of attitude has its greatest influence on behavior in the absence
of direct prior experience. In contrast, the linear relationship between
evaluations of blood donation and self-reported behaviors did not vary
as a function of prior experience. Thus an important difference between
affect and evaluation concerns their distinct roles in the attitude-to-behavior
process.
Study 2 helps to establish some conditions under which direct prior
experience does not increase the chronic accessibility of attitudes from
memory (cf. Fazio, 1986). Indeed, it may become less likely that attitudes
will be spontaneously retrieved from memory when behaviors become
 AFFECT VERSUS EVALUATION 269

habitual (cf. Ronis et al., 1989). The distinct influences of affect and
evaluation on behavior also change in unique ways as a function of prior
experience.
CONCLUSION
The present results support the distinction between affect and evaluation
in the structure of attitudes, This conclusion follows from a nomobgical
network approach to construct validation (Cronbach & Meehl, 1955).
Although each result taken in isolation provides equivocal evidence, the
entire set of results provides clear empirical confirmation of the theoretical
distinction.
Previous studies have indicated a very high correspondence between
affective and evaluative components of attitude. Indeed, McGuire (1968)
concluded that attitude components, such as affect and evaluation, “have
proven to be so highly intercorrelated that theorists who insist on dis-
tinguishing them should bear the burden of proving that the distinction
is worthwhile (p. 157).” The present results, along with those of previous
studies (Bagozzi, 1978; Breckler, 1984; Kothandapani, 1971; Ostrom,
B969), appear to indicate that affect and evaluation are, indeed, distin-
guishable components of attitude. It remains for future research to establish
the distinct roles played by affect versus evaluation in the structure of
attitudes.

REFERENCES
AIlen, M. J., & Yen, W. M. (1979). Introduction to measurement theory. Monterey, CA:
Brooks/Cole.
Bagozzi, R. P. (1978). The construct validity of the affective, behavioral, and cognitive
components of attitude by analysis of covariance structures. M&variate Behavioi-nl
Research, 13, 9-31.
Batra, R., & Ray, M. L. (1985). How advertising works at contact. In L. F. Alwitt & A.
A. Mitchell (Eds.), Psychological processes and advertising effects (pp. 13-43). H&dale,
NY: Erlbaum.
Breckler, S. J. (1983). Validation of affect, behavior, and cognition as distinct components
of attitude. Unpublished doctoral dissertation, Ohio State University, Columbus.
Breckler, S. J. (19S4). Empirical validation of affect, behavior, and cognition as distinct
components of attitude. Journal of Personality and Social Psychology, 47, 1191-1205.
Breckler, S. J., & Wiggins, E. C. (1989). On defining attitude and attitude theory: Once
more with feeling. In A. R. Pratkanis, S. J. Breckler, & A. G. Greenwald (&Is.),
Attitude structure and function (pp. 407-427). Hillsdale, NJ: Eribaum.
Buck, R. (1985). Prime Theory: An integrated view of motivation and emotion. Psycho@&/
Review, 92, 389-413.
Campbell, D. T., & Fiske, D. W. (1959). Convergent and discriminant validity by the
multitrait-multimethod matrix. Psychological Bulletin, 56, 81-105.
Chein, 1. (1951). Notes on a framework for the measurement of discrimination and prejudice.
In M. Jahoda, M. Deutsch, & S. W. Cook (Eds.), Research methods in social relations
(pp. 381-390). New York: Dryden Press.
Cohen, 3., & Cohen, P. (1983). Applied multiple regression/correlation analysis for Ihe
behavioral sciences (2nd ed.). Hi&dale, NJ: Erlbaum.
270 BRECKLER AND WIGGINS

Cronbach, L. J., & Meehl, P. E. (1955). Construct validity in psychological tests. Psychological
Bulletin, 52, 281-302.
Edwards, A. L. (1957). Techniques of attitude scale construction. New York: Appleton-
Century-Crofts.
Fazio, R. H. (1986). How do attitudes guide behavior? In R. M. Sorrentino & E. T. Higgins
(Eds.), Handbook of Motivation and Cognition (pp. 204-243). New York: Guilford
Press.
Fazio, R. H., & Zanna, M. P. (1981). Direct experience and attitude-behavior consistency.
In L. Berkowitz (Ed.), Advances in Experimental Social Psychology (Vol. 14, pp.
161-202). New York: Academic Press.
Fleming, D. (1967). Attitude: The history of a concept. Perspectives in American History,
1, 287-365.
Gorsuch, R. L. (1983). Facfor analysis (2nd ed.). Hillsdale, NJ: Erlbaum.
Greenwald, A. G. (1968). On defining attitude and attitude theory. In A. G. Greenwald,
T. M. Ostrom, & T. C. Brock (Eds.), Psychological foundations of attitudes (pp. 361-
388). New York: Academic Press.
Greenwald, A. G. (1982). Is anyone in charge? Personalysis versus the principle of personal
unity. In J. Suls (Ed.), Psychological perspectives on the self (Vol. 1, pp. 151-181).
Hillsdale, NJ: Erlbaum.
Hiigard, E. R. (1980). The trilogy of mind: Cognition, affection, and conation. Journal of
the History of the Behavioral Sciences, 16, 107-117.
Insko, C. A., & Schopler, J. (1967). Triadic consistency: A statement of aifective-cognitive-
conative consistency. Psychological Review, 74, 361-376.
Izard, C. E. (1972). Patterns of emotions: A new analysis of anxiety and depression. New
York: Academic Press.
Izard, C. E. (1977). Human emotions. New York: Plenum.
Katz, D. (1960). The functional approach to the study of attitudes. Public Opinion Quarterly,
24, 163-204.
Katz, D., & Stotland, E. (1959). A preliminary statement to a theory of attitude structure
and change. In S. Koch (Ed.), Psychology: A study of a science (Vol. 3, pp. 423-
475). New York: McGraw-Hill.
Kothandapani, V. (1971). Validation of feeling, belief, and intention to act as three components
of attitude and their contribution to prediction of contraceptive behavior. Journal of
Personality and Social Psychology, 19, 321-333.
Kramer, B. M. (1949). Dimensions of prejudice. Journal of Psychology, 27, 389-451.
Krech, D., & Crutchfield, R. S. (1948). Theory and problems of social psychology. New
York: McGraw-Hill.
Leventhal, H. (1984). A perceptual-motor theory of emotion. In L. Berkowitz (Ed.),
Advances in Experimental Social Psychology (Vol. 17, pp. 117-182). New York:
Academic Press.
Long, J. S. (1983). ConJirmatoryfactor analysis. Beverly Hills, CA: Sage.
Mann, J. H. (1959). The relationship between cognitive, affective, and behavioral aspects
of racial prejudice. Journal of Social Psychology, 49, 223-228.
McGuire, W, J. (1968). The nature of attitudes and attitude change. In G. Lindzey & E.
Aronson (Eds.), The handbook of social psychology (2nd ed., Vol. 3, pp. 136-314).
Reading, MA: Addison-Wesley.
McGuire, W. J. (1985). Attitudes and attitude change. In G. Lindzey & E. Aronson (Eds.),
Handbook of social psychology (3rd ed., Vol. 2, pp. 233-346). New York: Random
House.
Millar, M. G., & Tesser, A. (1986). Effects of affective and cognitive focus on the attitude-
behavior relation. Journal of Personality and Social Psychology, 51, 270-276.
Osgood, C. E., Suci, G. J., & Tannenbaum, P. H. (1957). The measurement of meaning.
Urbana, IL: Univ. of Illinois Press.
 AFFECT VERSUS EVALUATION 271

Ostrom, T. M. (1969). The relationship between the affective, behavioral and cognitive
components of attitude. Journal of Experimental Social Psychology, 5, 12-30.
&wall, R. M. (1977). A review of blood donor motivation and recruitment. Transfusion,
17, 123-135.
Piliavin, J. A., Callero, P. L., & Evans, D. E. (1982). Addiction to altruism? Opponent-
process theory and habitual blood donation. Journal ofPersonality and Social Psychology,
43, 1200-1213.
Piiiavin, J. A., Evans, D. E., & Callero, P. L. (1982). Learning to “Give to Unnamed
Strangers”: The process of commitment to regular blood donation. In E. Staub,
Bar-Tal, J. Karylowski, & J. Reykowski (Eds.), The development and maintenance
of prosocial behavior: International perspectives. New York: Plenum.
Ronis, D. L., Yates, J. F., & Kirscht, J. P. (1989). Attitudes, decisions, and habits as
determinants of repeated behavior. In A. R. Pratkanis, S. J. IBreckler, & A. 6.
Greenwald (Eds.), Attitude structure andfunction (pp. 213-239). Hi&dale, NJ: Erlbaum.
Rosenberg, M. J., Hovland, C. I., McGuire, W. .I., Abelson, R. P., & Brehm, J. W,
(Eds.). (1960). Attitude organization and change: An analysis of consistency among
attitude components. New Haven, CT: Yale Univ. Press.
Smith, M. B. (1947). The personal setting of public opinions: A study of attitudes toward
Russia. Public Opinion Quarterly, 11, 507-523.
Smith, M. B., Bruner, J. S., & White, R. W. (1956). Opinions Gndpersonality. New York:
Wiley.
Thurstone, L. L., & Chave, E. J. (1929). The measurement qfattitude. Chicago: Univ.
of Chicago Press.
Triandis, H. C. (1971). Attitude and attitude change. New York: Wiley.
Widaman, K. F. (1985). Hierarchically-nested covariance structure models for multitrait-
multimethod data. Applied Psychologica/ Measurement, 9, l-26.
Wilson, T. D., &Dunn, D. S. (1986). Effects of introspection on attitude-behavior consistency:
Analyzing reasons versus focusing on feelings. Journal of Experimenfal Socicrl Psg-
chology, 23, 249-263.
Woodmansee, J. J., & Cook, S. W. (1967). Dimensions of verbal racial attitudes: Their
identification and measurement. Journal ofpersonality and Social Psychology, ‘9,241~
250.
Zajonc, R. B. (1980). Feeling and thinking: Preferences need no inferences. American
Fsychologisr, 35, 1.51-175.
Zajonc, R. B., & Markus, H. (1984). Affect and cognition: The hard interface. In C. E.
Izard, J. Kagan, & R. B. Zajonc (Eds.), Emotions, cognition, and behavior (pp. 7%
102). Cambridge: Cambridge Univ. Press.