Journal of Educational Psychology Copyright 1986 by the American Psychological Association, Inc.

1986, Vol. 78, No. I, 52-58 0022-0663.'86/$00.75

Social Intelligence: Evidence for Multidimensionality and Construct

Independence
Herbert A. Marlowe, Jr.
Division of Continuing Education
Department of Counselor Education
University of Florida

The construct of social intelligence has been plagued by both definitional and psychometric
problems. The wide number of definitions in use indicate it may be a multidimensional
construct. However, psychometric evidence until recently has failed to support the hypothesis
that social intelligence is distinct from verbal intelligence. This study empirically derived a
multidimensional model of social intelligence. Through the use of factor analytic techniques,
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

five domains of social intelligence were identified. The five domains were found to be
This document is copyrighted by the American Psychological Association or one of its allied publishers.

independent of verbal and abstract intelligence. Areas of further investigation are indicated.

Educational and psychological researchers have hypothesized (concern for others), social self-efficacy, empathy skills (the abil-
for some time that a unique domain of abilities exists with respect ity to cognitively and affectively understand another), and social
to social content. This domain of abilities is referred to as social performance skills (observable social behaviors).
intelligence or social competence. Social intelligence is the ability The wide variety of social intelligence definitions indicates that
to understand the feelings, thoughts, and behaviors of persons, social intelligence may be a multidimensional construct. Scarr
including oneself, in interpersonal situations and to act appropri- (1981) acknowledged this possibility when she argued that intel-
ately upon that understanding. It is composed of a set of problem- lectual competence is inextricably entwined with motivational and
solving skills that enable the individual to find and/or resolve adjustment characteristics. An empirical investigation of the mul-
interpersonal problems and create useful social products (Mar- tidimensional alternative is therefore imperative if advances are to
lowe, 1985). Social intelligence may therefore be equated with be made in definitional clarity.
social competence. The second major problem for social intelligence researchers
With respect to definitional issues, the general field of intelli- has been the confounding of social intelligence and academic (ver-
gence is beset by a lack of a universally agreed upon definition bal and quantitative) intelligence (Keating, 1978; Walker & Foley,
(Green, 1981). With respect to social intelligence, five interrelated 1973). This confounding has meant that the distinctiveness and
but partly distinct definitions can be identified. One is a motiva- empirical coherence of the social intelligence domain has not been
tional definition that views competence as the organism's capa- well established. Since the construct of social intelligence was
bility for developing goals and then generating goal-directed activity. proposed in 1920 by Thorndike, attempts to measure social intel-
A closely related definition of competence is self-efficacy, that ligence have not been successful. Factor analyses of the early
is, an individual's expectation of personal mastery and success. social intelligence tests showed that they were heavily loaded with
Another definition of competence is the ability to perform actions verbal and memory-ability variance, with little evidence of any
that bring positive reinforcements or eliminate negative reinforce- unique variance that could be identified as "social intelligence."
ments; in other words, competence is a skill. A fourth way to Strieker (1982) reported that many of the later tests suffered from
define competence is to focus on performance. From this view- this same problem, that is, a substantial overlap with general in-
point the competent person is effective in performing socially tellectual ability. Although theoretically and intuitively reason-
beneficial behaviors, for example, achieving objectives, etc. Fi- able, a distinct realm of social intelligence has not been found.
nally, competence is defined as personality traits that are dem- In analyzing why, Keating (1978) raised the possibility that the
onstrated in organized patterns of cognitive, affective, and/or format of social intelligence measures activated an academic set.
behavioral functioning. Several researchers have developed con- Green (1981) also argued that verbal reasoning is pervasive among
ceptual models of social competence that combine two or more performance measures because of the use of paper and pencil
of these different approaches (Meichenbaum, Butler, & Gruson, measures. Keating also argued that social intelligence researchers
1981; O'Malley, 1977; Tyler, 1978; White, 1974). Marlowe (1984, had used too narrow a definition of the construct by failing to
1985) developed from the empirical literature a model of social emphasize social performance.
intelligence that comprised four major domains: social interest A major failing therefore of past social intelligence researchers
was their inability to establish the distinctiven^ss of social intel-
ligence. However, recent studies have improved on the method-
This study represents portions of the author's doctoral dissertation. ological problems of earlier studies and offer stronger support for
Committee members were Bert Sharp (chair), Ben Barger, and Larry Loesch.
the independence of the construct.
As well as to the committee, the author's appreciation is expressed to
Hardy Hall and Mark Prange for their statistical assistance.
Examining Keating's hypothesis, Marlowe and Bedell (1982)
Correspondence concerning this article should be addressed to Herbert investigated further the independence of the social intelligence
A. Marlowe, Jr., TREEO Center, 3900 SW 63rd Boulevard, University construct. As their measure of social intelligence they utilized the
of Florida, Gainesville, Florida 32608. Minnesota Multiphasic Pesonality Inventory (MMPI) Social In-

52
 SOCIAL INTELLIGENCE 53

troversion (SI) subscale. They selected the SI scale because it had less than 12th grade. All participants were employees in a state-
been found to be equivalent to a behavioral role-play as a predictor funded mental hospital and were randomly selected. Thirty percent
of social skill. Also, its true-false format minimized the need for were professionals and 70%, paraprofessionals.
academic ability. To determine the relation between social intel- As is evident from these descriptive statistics, the participants
ligence and general intelligence, Marlowe and Bedell correlated differed from the general population in being predominantly fe-
SI scores with Shipley-Hartford abstract and verbal intelligence male. Minority participants were also overrepresented in terms of
scores in a sample of 479 adult psychiatric clients. They found the general population. Both of these trends may be attributed to
nonsignificant correlations (r = — .003 and - .009, respectively). the job characteristics of the state mental hospital setting. The
On the basis of these results, Marlowe and Bedell argued that majority of positions in this system are low-level, minimal edu-
social intelligence is a distinct domain that may be masked by cational requirement positions that are most often filled by women
assessments relying on reading, writing, and other academic abil- and minorities.
ities.
A second study addressing this question was conducted by Ford
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

and Tisak (1983). They used four measures of academic intelli- Procedure and Measures
This document is copyrighted by the American Psychological Association or one of its allied publishers.

gence and six measures of social intelligence. Ford and Tisak, as

did Marlowe and Bedell, defined social intelligence as effective Administration of the measurements entailed dividing the as-
social performance. Ford and Tisak's measures of social intelli- sessment battery into two groups, A and B, and administering
gence were therefore selected to measure social performance. On them in a counterbalanced order. The following measurements
the basis of two samples totaling 620 persons, Ford and Tisak's were used:
results supported the position that social intelligence is a distinc- Social interest. The participants' levels of social interest were
tive domain. In terms of convergent-discriminant validity evi- assessed via the Social Interest Index (Greever, Tseng, & Fried-
dence, they found the social intelligence within-domain correlations land, 1973) and the Social Interest Scale (Crandall, 1975). The
of their two samples to be larger than the between-domains cor- Social Interest Index (SII) is a 32-item scale on which a high score
relations with academic measures. Through a factor analysis they indicates a higher level of social interest. The SII is designed to
also identified a social intelligence factor. Finally, a regression measure a person's willingness to contribute and cooperate within
analysis indicated that the four significant variables in the regres- the areas of four life tasks: work, friendship, love, and self-sig-
sion on a behavioral measure of social competence were social nificance. The Social Interest Scale (SIS) is a 24-item scale with
intelligence variables. These social intelligence variables ac- a forced-choice format. It is designed to measure a person's de-
counted for a substantial proportion of the variance (31%). The sired interest in the welfare of others. Although administered, the
remaining variables, including academic aptitude variables, could SIS was not used in the factor analysis because there was a highly
account only for an additional 1% of the variance. unbalanced split of its dichotomous responses. This imbalance
These two studies support the position that social intelligence made factor analysis inappropriate (Comrey, 1978).
is a different form of intelligence than is general academic (verbal Social self-efficacy. Two measures of social self-esteem were
and quantitative) intelligence. It is apparent that social intelligence used to assess self-efficacy in social situations. The Self-Efficacy
has been merged in the past with general intelligence because of Scale (Sherer et al., 1982) is a 36-item scale composed of two
the types of measures used and the operational definition of the factorially derived subscales, a general scale and a social scale.
construct. When social intelligence is defined in terms of social For this study only the latter was used, called the Social Self-
effectiveness and measured by instruments not confounded with Efficacy Scale (SSES). The SSES is a six-item scale on which a
academic ability requirements, the construct is distinct and iden- high score indicates high social self-efficacy. The Texas Social
tifiable. Behavior Inventory, Short Form A (TSBIA; Helmreich & Stapp,
This study addressed the two major issues facing social intel- 1974) is a 16-item short form of the 32-item TSBI. The TSBIA is
ligence researchers: examining the multidimensional nature of the a measure of self-esteem or perceptions and feelings of compe-
construct and testing if it is independent of general intelligence. tence in social situations. High scores indicate high levels of self-
Using Marlowe's four-part definition as a guide, we addressed the esteem.
following research questions: Empathy skills. Two measures of empathy skills were used to
1. What are the main domains of social intelligence? assess empathy skills. The Perceived Decoding Ability Scale, Form
2. To what extent are any domains of social intelligence in- 2 (PDA2; Zuckerman & Larrance, 1979) is a 16-item measure of
dependent of verbal and quantitative intelligence? a person's capability to decode or interpret the nonverbal cues of
emotion emitted by another person. A high score on the instrument
indicates a high level of decoding ability. The Interpersonal Reac-
Method
tivity Index (IRI; Davis, 1983) is a 28-item self-report measure
Subjects designed to assess four dimensions of empathy: perspective tak-
ing, fantasy, empathic concern, and personal distress. The per-
The 188 participants (12.8% male, 83.5% female, 3.7% no spective-taking (PT) scale is a seven-item scale that assesses the
report) ranged in age from 18 to 65, with a mean age of 43.4. tendency to spontaneously adopt the psychological point of view
Slightly more than 50% of the sample were members of a minority of others. The empathic concern (EC) scale is a seven-item scale
group (non-Caucasian), and more than 62% were married or co- that assesses "other-oriented" feelings of sympathy and concern
habiting. In terms of educational level, 59.4% of the participants for unfortunate others. In this study only these two scales were
had at least some college education, with 8.3% having completed used; the remaining two scales were not interpersonal in focus.
 54 HERBERT A. MARLOWE, JR.

These two scales, PT and EC, were selected because they were cedure was used to derive the first-order factors. Initial factor
specifically designed to measure the two dominant dimensions of selection was based on eigenvalues of 1.0. A second-order factor
empathy: cognitive empathy and affective empathy. analysis was then conducted, using the first-order factors, to de-
Social skills. Two measures of social skills were used to assess termine the factor structure of the overall instrumentation. Again,
this domain. The Perceived Encoding Ability Scale, Form 2 (PEA2; the same statistical procedure was used. This approach was se-
Zuckerman & Larrance, 1979) is a 16-item measure of a person's lected because the number of subjects was insufficient to conduct
ability to transmit nonverbal cues of emotion. Items are scored so a first-order factor analysis on all scales combined. The second-
that a high score on the PEA2 indicates a high level of encoding order factor analysis resulted in the derivation of independent do-
ability. The Social Skills Survey — Peer (SSSP; Marlowe & mains of social intelligence.
Weinberg, 1983) is a 15-item, 5-point Likert self-report on per- To test whether the derived domains of social intelligence were
ceived level of social skill in social situations involving peers. It independent of general intelligence, I conducted an orthogonal
has an internal consistency (Cronbach's alpha) of .93 and a test- factor analysis (principal factors with varimax rotation) using the
retest reliability of .79. Construct validity was established by mod- social competence domain scores and the Shipley-Hartford verbal
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

erate correlations with a number of other measures of social skill and abstract IQ scores. Loading of the IQ scales on a separate
This document is copyrighted by the American Psychological Association or one of its allied publishers.

as described in the source article. factor or factors from those on which the social domains loaded
Intelligence. The Shipley-Hartford Institute of Living Scale (SH; would argue for the independence of the social intelligence do-
Shipley, 1940) measures verbal and abstract-thinking abilities. mains. Second, we computed a partial correlation matrix with the
Each portion of the scale has a 10-min time limit. The vocabulary sum scores of the social intelligence domain measures with the
test is composed of 40 multiple-choice items, and the abstract- variance accounted for by IQ removed. This matrix was computed
thinking test contains 20 completion items. for the two general IQ scores separately using a "Partial Corr"
Demographic questionnaire. A brief demographic questionnaire SPSS procedure. The partial correlations were compared against
was used to attain data on age, sex, educational level, marital their respective whole correlations to assess the significance of
status, and job position. the removal of IQ variance. Evidence for the independence of
social competence from general intelligence would result if the
correlations between the social intelligence domains remained es-
Statistics sentially the same after the effects of general IQ were removed.

In order to examine the multidimensional nature of social in- Results

telligence, factor analyses were conducted on each instrument to
Table 1 displays the means and standard deviations of partici-
determine the factoral structure of each scale. An orthogonal prin- pants' scores on the instruments used in the study as well as a
cipal factors with iteration and varimax rotation (SPSS PA2) pro- comparison of the present scores with those reported in other
studies. Participants in the present study were slightly higher in
Table 1 social interest, which may be attributed to their occupational choice.
Means and Standard Deviations of Participants' Scores on They were also lower in verbal intelligence. With these two ex-
Social Intelligence Measures and Relation of These Scores to ceptions, this sample did not differ from other samples reported
Comparison Group's From Source Articles in the literature.
Relation to The first question addressed in this study was what the internal
comparison domain structure of social intelligence was and how the derived
Instrument N M SD group"
domain structure compared with the hypothesized domain struc-
Social Interest Index 172 135.96 12.6 >(p < .01)
ture, consisting of the following four constructs; social interest,
Social Self-Efficacy Scale 187 23.55 5.86 nd
Texas Social Behavior social self-esteem, social cognition abilities, and social skill or
Inventory, Short Form A 184 59.37 8.4 >{p < .01) behavioral repertoire. As decribed in the Method section, this
Interpersonal Reactivity Index question was addressed through the use of first- and second-order
Empathic Concern 185 29.54 4.63 dna factor analyses. Table 2 presents the results of the second-order
Perspective Taking 183 28.21 4.43 dna factor analysis. The initial factor extraction identified six factors
Perceived Decoding 187 18.99 11.68 nd
Ability Scale, Form 2 with eigenvalues greater than 1.0. On the basis of this finding,
Perceived Encoding 186 12.54 11.05 nd one- through six-factor solutions were calculated. The rotated six-
Ability Scale, Form 2 factor matrix was rejected because of numerous splits and limited
Social Skills interpretability. Of the remaining alternatives, the five-factor so-
Survey — Peer 186 58.85 8.72 nd
lution was selected as the most logical and interpretable on the
Shipley-Hartford
Institute of Living Scale basis of eigenvalues and a scree test. As Table 2 indicates, this
Abstract 188 20.87 9.55 nd solution accounted for 41.2% of the variance after rotation. Al-
Verbal 187 25.17 6.07 <(p < .001) though Factor 1 accounted for a greater percentage of variance
a
t tests were calculated on each measure to determine if the mean score than did any of the other factors (13.6%), it did not account for
for this sample was significantly higher or lower than that of mean scores such a disproportionate amount of variance that it could be called
reported in source articles. A > indicates the mean for this sample is
higher; < indicates the mean is lower; nd indicates no difference; and dna
a general factor. Rather, these results indicate the presence of five
indicates that the data to make a comparison were not available. common factors.
 SOCIAL INTELLIGENCE 55

Table 2 It constitutes 12.8% of the variance on social intelligence, with

Second-Order Factor Analysis: Factor Structure With Loadings an eigenvalue of 2.18. These three first-order factors represent
Factor
social performance skills that can be behaviorally observed. For
Instrument this reason this factor is called Social Skills.
factor 1 2 3 4 5
The third factor extracted in this solution is composed of Factors
SII
2 .70 .16 .08 .17 .11 1 and 2 of the Perceived Decoding Ability Scale, Form 2 (PDA2).
3 .68 .09 .12 .24 .19 Because the PDA2 measures empathic ability, this factor clearly
TSBIA represents empathy skills and is so labeled: Empathy Skills ac-
1 .63 .20 .14 .19 .08 counts for 6.8% of the variance and has an eigenvalue of 1.15.
SII
The fourth factor extracted from the analyses comprises the two
1 .47 .03 .25 .06 .16
SSES factors from the Perceived Encoding Ability Scale, Form 2 and
1 .38 .11 .15 .19 .31 the first factor from the Interpersonal Reactivity Index. This factor
SSSP accounts for 4.5% of the variance, with an eigenvalue of .77. An
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

1 .17 .84 .23 .26 .07 examination of these items reveals that they all refer to either
This document is copyrighted by the American Psychological Association or one of its allied publishers.

3 .20 .78 .13 .06 .05

2 .18 .64 .21 .24 .14 emotional expressiveness or high sensitivity to other persons' af-
PDA2 fective states. As such, Factor 4 represents an apparent personality
2 .04 .24 .75 .15 .18 trait that was not hypothesized in the original model. Because this
1 .10 .26 .62 .20 .16 personality trait describes the degree to which a person is simul-
PEA2
1 .13 .14 .25
taneously emotionally expressive and sensitive, it is labeled Emo-
.22 .60
IRI tionality. The argument that this factor is a personality trait is
1 .13 .07 .09 .48 .18 supported by the fact that when the Social Interest Scale was
PEA2 included in the second-order factor analysis for purposes of con-
2 .25 .19 .21 .48 .26 ceptual clarification, it loaded on this factor throughout various
SSES
1 .22 .08 .01 .11 .53 solutions. The Social Interest Scale asks respondents to describe
IRI their personality by selecting adjectives.
1 .06 .19 .14 .15 .47 The fifth factor extracted in the solution comprises the first-
TSBIA
order factors from the Texas Social Behavior Inventory, the In-
2 .18 .12 .14 .21 .47
terpersonal Reactivity Index, and the Self-Efficacy Scale. This
Summary statistics (after rotation) factor accounts for 3.5% of the variance, with an eigenvalue of
Eigenvalue 2.31 2.18 1.15 .77 .60 .60. An examination of the items reveals that they refer to a lack
% of variance 13.6 12.8 6.8 4.5 3.5 of confidence and the presence of anxiety in the social situations.
Cumulative variance 13.6 26.4 33.2 37.7 41.2
In the two- or three-factor solutions, Factor 5 loaded with Factor
Note. SO = Social Interest Index (Greever, Tseng, & Friedland, 1973); 1, indicating the presence of an efficacy dimension in the factor.
TSBIA = Texas Social Behavior Inventory, Short Form A (Helmreich &
Stapp, 1974); SSES = Social Self-Efficacy Scale (Sherer et al., 1982); However, because Factor 1 includes both social interest and social
SSSP = Social Skills Survey — Peer (Marlowe & Weinberg, 1983); self-efficacy, it is conceptually clearer to distinguish these two
PDA2 = Perceived Decoding Ability Scale, Form 2 (Zuckerman & Lar- factors. For this reason, Factor 5 is called Social Anxiety.
rance, 1979); PEA2 = Perceived Encoding Ability Scale, Form 2 (Zuck- In review, the second-order factor analysis yielded five inter-
erman & Larrance, 1979); Interpersonal Reactivity Index (Davis, 1983).
pretable factors. Two of these factors, Social Skills and Empathy
Skills, are the same as the hypothesized domains of behavioral
Factor 1 is composed of first-order factors from the Social In- repertoire and social cognition. The Prosocial Attitudes factor rep-
terest Index (Factors 3 and 2), the TSBIA (Factor 1), and the SSES resents a combination of two of the hypothesized domains, social
(Factor 1) and constitutes 13.6% of the variance, with an eigen- interest and social self-efficacy. Because the first-order factors
value of 2.31. Examination of the individual items and the scales composing Factor 1 always loaded together, even in more complex
composing this factor indicates that it measures level of interest factor solutions, these results strongly indicate that social interest
in and concern for other people as well as confidence and enjoy- and social self-efficacy are not separate or distinctive domains.
ment of social situations. On the basis of this content analysis, it Instead, they represent a positive attitude toward oneself and toward
may be concluded that Factor 1 can be called Prosocial Attitude. others. Factor 5 represents a social anxiety domain that appears
Because this Prosocial Attitude represents both a level of interest to be the obverse of self-efficacy. As such it represents a modi-
in and concern for others as well as a sense of social self-efficacy, fication of the hypothesized domains. Factor 4, Emotionality, rep-
it may be argued that interest and efficacy are closely linked. This resents a dimension of social intelligence that was not originally
argument is supported by the fact that Factor 1 retained the same hypothesized. One dimension of social intelligence, therefore, ap-
structure in the more fragmented six-factor solution. On the basis pears to be emotional expressiveness and sensitivity.
of this, one dimension of social intelligence can be defined as The second question this study attempted to answer was whether
Prosocial Attitude. The factor combines both social interest and social intelligence was a separate domain from general intelli-
social self-efficacy. It represents a positive perception of social gence. To answer this question two analyses were conducted.
interaction. First, a factor analysis was used to determine the factor pattern
The second factor of the second-order analysis is formed solely of the five social intelligence domains and the two general intel-
of the three factors derived from the Social Skills Survey — Peer. ligence measures. Second, social intelligence correlation matrices
 56 HERBERT A. MARLOWE, JR.

Table 3 are significantly correlated with Prosocial Attitude (r = .31, - 2 2 ,

Rotated Factor Analysis for General and Social Intelligence respectively).
Measures The position that general intelligence is distinct from social
Factor intelligence was supported in all cases except one. With this one
Measure 1 2 3
exception all correlations were low and nonsignificant. The one
significant relation was between Social Anxiety and the Shipley-
Shipley-Hartford
Verbal Scale .87 .23 .16 Hartford Verbal Scale (r = - . 1 9 , p < . 0 1 ) . Although this re-
Shipley-Hartford lation is not strong, it does indicate that verbal facility is related
Abstract Scale .65 .13 .20 to anxiety in social situations. In contrast to these low correlations,
Prosocial Attitude .12 .89 .07 the relation between the two general intelligence measures, ab-
Emotionality .19 .34 .07 stract and verbal, was strong and significant (r = .59, p < .0001).
Social Anxiety .09 .28 .05
Social Skills .11 .24 .73 The hypothesis that general and social intelligence are indepen-
Empathy Skills .12 .20 .39 dent domains is further supported by the partial correlation ma-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

trices reported in Table 5. The Shipley-Hartford Verbal and Abstract

This document is copyrighted by the American Psychological Association or one of its allied publishers.

Summary statistics (after rotation)

Eigenvalue 1.22 .98 .67
scores were partialed out of the correlations of the social intelli-
% of variance 17.6 14.0 9.6 gence measures. In neither case did removal of the Shipley score
Cumulative change in any significant way the correlation pattern with respect
variance 17.6 31.6 41.3 to direction or strength. This indicates that the relation between
the social intelligence domains is the same regardless of the pres-
ence or absence of general intelligence. In other words, general
were calculated with and without general intelligence partialed intelligence does not play a significant part in this relation.
out. The difference, if any, between full and partial correlation
matrices would indicate the relative importance of general intel-
ligence. Discussion
The initial factor extraction yielded three factors with eigen-
values greater than 1.00. In accord with the factor analysis pro-
cedures used to examine multidimensionality, one- and two-factor Multidimensionality
solutions were also calculated. Of the three calculated solutions,
the three-factor solution was selected as the most interpretable. The first question this study sought to examine was the multi-
Table 3 displays the results of the three-factor solution, indicating dimensionality of social intelligence. The results of the second-
that the first factor is composed of the two general intelligence order factor analysis reported above yielded five separate factors
measures. The second factor is composed of the Prosocial Attitude or domains: Prosocial Attitude, Social Skills, Empathy Skills,
and Emotionality factors, with Social Anxiety loading less than Emotionality, and Social Anxiety. These findings support the ar-
.30 on this factor. This second factor is clearly composed of social gument that social intelligence is multidimensional, as advanced
self-perception scales. Factor 3 is composed of the two skill fac- by Ford (1982) and Gerdeman (1975). However, Marlowe's hy-
tors: Social Skills and Empathy Skills. In sum, General Intelli- pothesized model was only partially supported. The skills domains
gence, Social Self-Perception, and Social Skills are three separate were empirically derived as theorized. Social interest and social
factors. On the basis of this analysis, General Intelligence is clearly self-efficacy appear to be one domain instead of two, as originally
a separate domain. However, social intelligence does not appear hypothesized. Furthermore, two unexpected domains, Social Anx-
to constitute a single domain but rather two distinct domains. iety and Emotionality, were derived. Each of these unexpected
The zero-order correlation analysis is shown in Table 4. The findings is discussed below.
correlation pattern supports the factor analysis reported earlier in With regard to social interest and social self-efficacy, because
that the two skills domains are significantly correlated with each the two constructs have developed from different origins (social
other (r = .34). Furthermore, Emotionality and Social Anxiety interest from the work of Alfred Adler and his associates; social

Table 4
Zero-Order Correlation Matrix for Social and General Intelligence Measures
Measure
Prosocial Social Empathy Social Shipley-Hartford Shipley-Hartford
Measure Attitude Skills Skills Emotionality Anxiety Verbal Scale Abstract Scale
Prosocial Attitude 1.00 .06 .03 31*** -.22** -.04 -.04
Social Skills 1.00 34*** -.10 -.07 -.02 -.02
Empathy Skills 1.00 -.04 -.06 -.03 -.04
Emotionality 1.00 -.04 .07 .04
Social Anxiety 1.00 -.19* -.09
Shipley-Hartford Verbal Scale 1.00 .59***
Shipley-Hartford Abstract Scale 1.00
*p < .01. **p < .001. ***p <.0001.
 SOCIAL INTELLIGENCE 57

Table 5 ness. This factor was not hypothesized in the original model and,
Correlation Matrix for the Social Intelligence Measures, With the unlike the Prosocial Attitude and Social Anxiety factors, does not
Shipley-Hartford Verbal and Abstract Scales Partialed Out represent a combination of originally hypothesized constructs, like
Measure Prosocial Attitude does, or a variant of a component of the original
model, like Social Anxiety does. Although obviously possessing
Prosocial Social Empathy Emotion- Social
Measure Attitude Skills Skills ality Anxiety some dimensions of empathy, it did not load with the Empathy
Prosocial Skills factor. Instead it represents a new and additional compo-
Attitude 1.00 .06 .03 .33*(.32) ,25*(.24) nent. Unlike the other four factors, Emotionality seems to rep-
Social Skills 1.00 .34* -.10 .08 resent a personality pattern or, in Buss's (1980) terms, a temperament
Empathy Skills 1.00 .04 .07 or innate personality disposition. This conclusion is based on the
Emotionality 1.00 .03
nature of the scales composing the domain items and the fact that
Social Anxiety 1.00
the Social Interest Scale, which is comprised of personality de-
Note. Abstract scale values are in parentheses if different from Verbal
scriptors, loaded with this factor when it was included in the
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

scale in either value or significance level.

This document is copyrighted by the American Psychological Association or one of its allied publishers.

*p < .0001. analysis. One facet of social intelligence, therefore, appears to be

a full or rich affective life, as opposed to a truncated affective
experience.
self-efficacy from Ruth Wylie, Albert Bandura, and their asso- These results have several implications for researchers and gen-
ciates), there has been little theoretical discussion or empirical erate further areas to be explored. First, the data strongly support
research regarding the degree of overlap between these two con- a multidimensional model of social intelligence. Social intelli-
structs. This available empirical literature also indicates that these gence clearly is not a unitary construct and should not be ap-
two constructs significantly overlap. For example, Mozdzierz and proached as if it were in research studies. Rather, there appears
Semyck (1980) found that persons high in social interest were also to be at least five distinct domains of social intelligence that al-
high in success motivation, whereas persons low in social interest though related are also empirically and conceptually distinguish-
were more concerned with failure avoidance. Crandall (1975) and able.
Paloutzian, Jackson, and Crandall (1978) found social interest Several points need further study. First, for these conclusions
significantly and positively correlated with perceived meaning- to be more tenable, replication studies should be performed with
fulness of life. Crandall and his colleagues (Crandall & Kytonen, similar and differing populations. Only as these domains are con-
1980; Crandall & Putnam, 1980) also found a significant positive sistently replicated can there be a high level of confidence that
relation between level of social interest and several measures of
these domains do indeed constitute social intelligence. Second,
psychological well being.
the search for additional domains of social intelligence should
Crandall (1978) offered a theory that may explain the relation continue. Third, there is the need to examine in greater detail,
of these two constructs. He argued that valuing, that is, to have through additional studies, some of the unexpected outcomes of
an interest or to care about something, is the fundamental aspect this study. For example, is Prosocial Attitude a coherent domain,
of psychological functioning that integrates other dimensions. If as this study indicates, or is it composed of other factors? Are
social interest is construed as the valuing of others and self-effi- there additional elements of social anxiety which should be in-
cacy is the valuing of oneself, then a common linkage is identified. corporated in the model? Finally, further analyses such as path
It may well be that value is the underlying dynamic resulting in and cluster analytic studies are needed to determine the interac-
this single factor of prosocial attitude. tions and relations among these components.
Another factor derived from the data, Social Anxiety, was not
directly hypothesized in Marlowe's original model. However, in-
asmuch as the factor comprises negatively worded self-efficacy Relation of Social Intelligence to General Intelligence
items, it apparently is the obverse of self-efficacy. An inverse
relation between social anxiety and self-efficacy has been re- The second question of this study asked what the relation of
ported. Nichols (1974), for example, found that psychiatric pa- social intelligence is to general or academic intelligence. The gen-
tients characterized by severe social anxiety had low self-esteem. esis of this question was the position of some researchers that
As such, social anxiety represents a lack of confidence in oneself social intelligence does not exist as a separate domain distinct
in social situations and the presence of discomfort in the presence from academic intelligence.
of others. This discomfort is the distinguishing characteristic of The data reported in the results section clearly reject this ar-
social anxiety, according to Buss (1980). As the factor analysis gument. In the present factor analysis, the social intelligence do-
indicates, there is a distinctive dimension from Prosocial Attitude, mains were distinguishable from the general intelligence domains.
which also contains elements of self-efficacy. Just as the Prosocial With one exception, there were nonsignificant correlations be-
Attitude domain tapped the dynamic of social value, the Social tween social intelligence and general intelligence. The magnitude
Anxiety factor appears to clearly tap the dimension of felt potency. of this one correlation was not particularly strong (r = - .19).
Although one may value oneself and others, this valuing is no Stronger evidence of the independence of social and general in-
guarantee that one will perceive oneself as potent in a social sit- telligence was provided by the results of the partial correlation
uation or that one will not feel discomfort in social situations. analysis.
Another factor extracted in the solution, Emotionality, was It is clear from this data that social intelligence and general
composed of items related to affective sensitivity and expressive- intelligence are separate domains. However, there is the possibil-
 58 HERBERT A. MARLOWE, JR.

ity that these results may be due to method variance in that self- H. A. Marlowe & R. B. Weinberg (Eds.), Competence development:
report measures were compared with performance measures. Fur- Theory and practice in special populations (pp. 50-52). Springfield,
ther studies however, should use other measures of general intel- IL: Charles C Thomas.
ligence and a full multitrait-multimethod design for this conclusion Marlowe, H. A., & Bedell, J. R. (1982). Social intelligence: Further
evidence for the independence of the construct. Psychological Reports,
to be confirmed.
51, 461^(62.
Marlowe, H. A., & Weinberg, R. B. (1983). The Social Skills Survey:
Construction and validation of the survey. Unpublished manuscript,
References
University of South Florida, Florida Mental Health Institute, Tampa,
FL.
Buss, A. H. (1980). Self-consciousness and social anxiety. San Francisco,
Meichenbaum, D., Butler, L., &Gruson, L. (1981). Toward a conceptual
CA: Freeman.
model of social competence. In J. D. Wine & M. D. Smye (Eds.),
Comrey, A. L. (1978). Common methodological problems in factor an-
Social competence (pp. 37-53). New York: Guilford.
alytic studies. Journal of Consulting and Clinical Psychology, 46, 648-
Mozdzierz, G. J , & Semyck, R. W. (1980). Relationship between al-
659.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

coholics' social interest and attitude toward success and failure. Journal
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Crandall, J. E. (1975). A scale for social interest. Journal of Individual

of Individual Psychology, 36, 61-65.
Psychology, 31, 187-195.
Nichols, K. A. (1974). Severe social anxiety. British Journal of Medical
Crandall, J. E. (1978). Social interest: A reply to Bickhard and Ford and
Psychology, 47, 301-306.
an alternative formulation. Journal of Individual Psychology, 34, 11-
O'Malley, J. M. (1977). Research perspective on social competence. Mer-
26.
rill-Palmer Quarterly, 23, 29-44.
Crandall, J. E., & Kytonen, J. A. (1980). Sex, age and stress as mod-
Paloutzian, R. F., Jackson, S. L., & Crandall, J. E. (1978). Types of
erators for the relation between social interest and well-being. Journal
Christian belief system and conversion experience as related to purpose
of Individual Psychology, 36, 169-182.
in life, social interest, religious orientation, and dogmatism. Journal of
Crandall, J. E., & Putnam, E. L. (1980). Relations between measures of Psychology and Theology, 6, 266-275.
social interest and psychological well being. Journal of Individual Psy-
Scarr, S. (1981). Testing for children: Assessment and the many deter-
chology, 36, 156-168.
minants of intellectual competence. American Psychologist, 36, 1159-
Davis, M. H. (1983). Measuring individual differences in empathy: Evi-
1166.
dence for a multidimensional approach. Journal of Personality and So-
Sherer, M., Maddux, J. E., Mercandante, B., Prentice-Dunn, S., Jacobs,
cial Psychology, 44, 113-126.
B., & Rogers, R. W. (1982). The Self-Efficacy Scale: Construction
Ford, M. (1982). Competence: Current conceptual and empirical status
and validation. Psychological Reports, 51, 663-671.
and implications for mental health. Unpublished manuscript, Stanford
Shipley, W. C. (1940). A self-administering scale for measuring intellec-
University, Palo Alto, CA.
tual impairment and deterioration. Journal of Psychology, 9, 371-377.
Ford, M., & Tisak, M. S. (1983). A further search for social intelligence.
Strieker, L. J. (1982). Interpersonal Competence Instrument: Development
Journal of Educational Psychology, 75, 197-206.
and preliminary findings. Applied Psychological Measurement, 6, 69-
Gerdeman, E. M. (1975). The contribution of social intelligence to pre-
81.
dicting accuracy of interpersonal perception. Dissertation Abstracts In-
Tyler, F. B. (1978). Individual psychosocial competence: A personality
ternational, 36, 470B. (University Microfilms No. 75-14509)
configuration. Educational and Psychological Measurement, 38, 309-
Green, B. F. (1981). A primer of testing. American Psychologist, 36,
323.
1001-1011.
Walker, R. E., & Foley, J. M. (1973). Social intelligence: Its history and
Greever, K. B., Tseng, M. S., & Friedland, B. V. (1973). Development
measurement. Psychological Reports, 33, 451^159.
of the Social Interest Index. Journal of Consulting and Clinical Psy-
White, R. C. (1974). Strategies of adaptation: An attempt at systematic
chology, 41, 454-458.
description. In G. V. Coehlo, D. A. Hamburg, & J. E. Adams (Eds.),
Helmreich, R., & Stapp, J. (1974). Short forms of the Texas Social Be-
Coping and adaptation (pp. 37-53). New York: Basic Books.
havior Inventory (TSBI): An objective measure of self-esteem. Bulletin
Zuckerman, M., & Larrance, D. T. (1979). Individual differences in per-
of the Psychonomic Society, 4, 473-475.
ceived encoding and decoding abilities. In R. Rosenthal (Ed.), Skill in
Keating, D. K. (1978). A search for social intelligence. Journal of Edu-
nonverbal communication: Individual differences (pp. 170-195). Cam-
cational Psychology, 70, 218-233.
bridge, MA: Oelgeschlager, Gunn & Hain.
Marlowe, H. A. (1984). The structure of social intelligence. Unpublished
doctoral dissertation, University of Florida, Gainesville, FL. Received April 25, 1984
Marlowe, H. A. (1985). Competence: A social intelligence approach. In Revision received September 17, 1985 •