Psychological Reports, 1974, 35, 1059-1070.

@ Psychological Reports 1974

DIMENSIONALITY OF THE ROTTER

INTERPERSONAL TRUST SCALE

KI-TAEK CHLTN AND JOHN B. CAMPBELL1

Institute for Social Research, University of Michigan

Summary.-The Rorter Interpersonal Trust Scale was interspersed in a larger

questionnaire completed by 187 paid undergraduates. Ics dimensionality was
investigated through cluster analysis and orthogonal and oblique rotations fol-
lowing a principal axis factor solution. Four dimensions, designated "Political
Cynicism," "Interpersonal Exploitation," "Societal Hypocrisy," and "Reliable
Role-performance" were identifiable across techniques. This finding of multi-
dimensionality underscores the incongruence between Romer's definition and
his measure of interpersonal trust. Studies supporting the multidimensional
interpretation are discussed. Scoring these four dimensions in addition to the
total scale is recommended. A short form of the sca!e consisting of the three
marker items for each dimension is proposed. This short form is efficient, and
it retains the dimensionaXty of the total scale.

The topic of trust is a serious social and political issue in these days of
"credibility gaps," one which lies at the core of our democratic society. Despite
its obvious relevance, the topic has been generally neglected in psychological
research. Fifteen years ago, Deutsch (1958) noted the inattention of psy-
chologists to the phenomenon of trust. This lack of attention persisted, and
Giffin ( 1967) later renewed the plea for sy stematic research. Recently, how-
ever, there are indications of increasing interest in crust among psychologists
(see Rotter, 1971; Singer & Singer, 1972, pp. 397-398).
When research interest in an area gathers momentum, as in the area of uust,
the initial high enthusiasm often results in insufficient attention to both the
refinement of measures and the development of a guiding theory. In the name
of continuity of research, pressure then develops for the continued use of exist-
ing measures, without further examination of the measures' operating charac-
teristics or theoretical bases. When a measure is either less than fully refined
or not linked with a substantive theory, the data which accumulate from its con-
tinued use are likely to be coarse in quality and equivocal in meaning. The
accumulation of studies in turn increases the pressure for further use of the
same measures, thus completing a spiraling cycle.
Of the three current lines of research on trust within psychology, the one
that stems from Rotter's ( 3 6 7 ) Interpersonal Trust Scale has been most active
'This study was s u p r d in p n by b e National Science Foundation's Gpability-build-
ing Grant GI-2990 to the Iustimre for Social Research, and by a grant from the Inter-
national Liason Committee for Research on Korea. Tohn R. P. French. Tr.. Sidney Cobb.
Robert D. Caplan, and Samuel R. Pinneau, Jr. have provided helpful commentaries on an
earlier version of this paper.
1060 K-T. CHUN & J. B. CAMPBELL

(Rotter, 1971).? W e suspect that the kind of spiraling cycle described above
is developing around the Interpersonal Trust Scale. Because of the growing
prominence of the Interpersonal Trust Scale as a research cool in the area of
trust, it seems desirable to examine its properties.
Rotter (1967) defines interpersonal trust as "an expectancy held by an
individual or a group that the word, promise, verbal or written statement of
another individual or group can be relied upon" (p. 652). Guided by this
definition, he has developed the Interpersonal Trust Scale as "an additive scale
in which a high score would show trust for a great variety of social objects such
as parents, teachers, physicians, politicians, classrnaces, friends, etc." (p. 653).
However, several of the items comprising the Interpersonal Trust Scale appear
to have questionable relevance to Rotter's definition of interpersonal trust3
Furthermore, our examination of the items raised questions regarding the dimen-
sionality and the efficiency, i.e., the presence of redundant or ineffective items,
of the Interpersonal Trust Scale. The present study examines these issues of
dimensionality and efficiency.
METHOD
The 25-item Interpersonal Trust Scale (that is, the 15 filler items were
deleted) was interspersed in a larger questionnaire which 187 undergraduates
(71 males and 116 females) were paid to complete. The questionnaire follows
the Rotter scale in using a five-point Likerc format with response categories of
( 1 ) "Strongly agree," ( 2 ) "Mildly agree," ( 3 ) "Agree and disagree equally,"
( 4 ) "Mildly disagree," ( 5 ) "Strongly disagree." Consequently, scores on the
Interpersonal Trust Scale can range from 25 to 125. Responses to the 12 items
worded in a "trustful" direction were reversed in the scoring procedure, such
that higher scores on the scale are indicative of high trust. The questionnaire
was administered in small groups (median size = 10, range = 4 to 22) during
the 2-wk. period directly preceding the 1972 Presidencial election.
T h e other two research lines are represented in the works of Erikson and Wrightsman.
Erikson's (1950, 1968) well-known use of the notion "basic trust vs basic mistrust" as
descriptive of the first stage in his eight-stage conception of human development has yet
to be translated into research idioms that fully embody rhe richness of the theory. Con-
sequently, his insights into trust have had little impact on (empirical) research [see
Constantino le (1967, 1770) for an attempt at operationalizing the Eriksonian scheme of
developmen$. Wrightsman (1964) has developed a Philosophies of Human Nature
scale which contains a "Trustworthiness" subscale. Although its heuristic value has been
impressive, the Trustworthiness subscale is open to criticism as a measure of trust (K. T.
Chun & J. B. Campbell, Notes on the internal structure of Wrightsman's measure of
trustworthiness. Unpublished manuscript, Univer. of Michigan, 1973).
=TOconfirm this impression, nine judges (advanced graduate students and post-doctoral
staff) were given Rotter's definition and the stated purpose of the Interpersonal Trust
Scale and were asked to rate on a five-point scale the face validity of each item. Over
half of our judges rated 1 0 of the 25 items to have either debatable or no face validity with
regard to Rotter's definition of interpersonal trust. For example. all judges agreed that
the item, "The future seems very promising," was irrelevant and that they would not have
included it in the scale.
 DIMENSIONALITY OF ROTTER INTERPERSONAL TRUST 1061

In order to investigate the dimensionality of the Interpersonal Trust Scale,

two techniques were applied to the inter-item correlation matrix based on the
total sample. First was the ICLUST cluster analysis r ~ u t i n e . ~This technique
forms clusters of icems on the basis of item proximity indices, continuing to add
items into clusters and to group clusters into higher order clusters as long as the
internal consistency (coefficient alpha) is increased by the addition. Second
was a principal axis factor analysis of the correlation matrix (with squared
multiple correlations inserted as communalicy estimates). This was followed
by both orthogonal (normalized Varimax) and oblique (normalized Biquar-
timin) rotations of 10 factors (since a principal components analysis of the
original correlation matrix with unities in the diagonal identified 1 0 factors with
latent roots h 1.0).
In addition, the REDUCE routine (Chun & Pinneau, 1973) was utilized
to assess the efficiency of the Interpersonal Trust Scale, i.e., the degree to which
there is a lack of redundant icems. This routine identifies the order in which
items can be combined to form subsets of increasing size which are optimal in
terms of part-whole correlations, and which maintain the dimensionaliry of the
original scale.
RESULTSAND DISCUSSION
The mean score of the Interpersonal Trust Scale for our sample was 66.86
(SD = 10.66). This value is very close to the most recent mean score reported
in the literature for a college sample, the figure of 66.64 (SD = 9.69) obtained
by Hochreich and Rotter (1970) in September, 1969. The mean scores for our
male (M = 68.17, S D = 9.85) and female ( M = 66.05, S D = 11.03) Ss did
not differ significantly ( t = 1.35, d f = 185, p > .15).
The cluster analysis indicated the presence of four clusters. The first
cluster was comprised of four items (4, 13, 16, 40), the second of eight items
(3, 8, 11, 14, 26, 31, 34, 37), the third of four icems (2, 18, 21, 29), and the
fourth of three items (6, 32, 36) (see Table 1 for the individual icems and their
loadings). Intercorrelatiocs among the cluscers ranged from .23 (second with
third) to .44 (first with third), with a median of 31.
The four clusters were compared against each one of the 10 orthogonal
and 10 oblique factors in terms of loadings of individual icems. This comparison
identified four of the orthogonal factors that matched the four clusters. The
"Orthogonal factors" column of Table 1 shows the factor loadings of the marker
icems (those with loadings k .30) for these four factors, which ranked first,
second, third, and seventh in proportion of variance accounted for (accounting
altogether for 47% of the common variance). Three of the oblique factors
showed a striking match to Clusters I, 11, and 111. These three factors ranked
'J. A. Kulik, W. R. Revelle, & C. C. Kulik. Scale construction by hierarchical cluster
analysis. (Unpublished manuscript, Univer. of Michigan, 1970)
 K-T.CHUN & J. B. CAMPBELL

TABLE 1
CLUSTER AND F A c r o ~LOADINGS OF THE ROlTER INTERPERSONAL
TRUST SCALE ITEMS
Items Clusters
Orthogonal Oblique Reducec
factors factors
6 Ca Cs G VI V3 Va Vd Ba
4. "This country has a dark future
unless we can attract
better people into politics. 52" -46 44 43
13. The judiciary is a place where we
can all get unbiased treatment. -49 50 -58
40. If we really knew what was going on
in international politics, the public
would have more reason to be fright-
ened than they now seem to be. 57 -47 53
16. Most people would be horrified if
they knew how much news the pub-
lic hears and sees is distorted. 47 31
3. In dealing with strangers one is better
off to be cautious until they have
provided evidence that they are
trustworthy. 58
14. It is safe to believe that in spite of
what people say, most people are
primarily interested in their own
welfare. 52
26. In these competitive rimes one has to
be alert or someone is likely to take
advantage of you. 46
8. Using the Honor System of not hav-
ing a teacher present during exams
would probably result in increased
cheating. 33
11. Most people can be counted on to do
what they say they will do. -40
31. Most idealists are sincere and usually
practice what they preach. -30
34. Most students in school would not
cheat even if they were sure of getting
away with it. -47 -34-32
37. A large share of accident claims filed
against insurance companies are
phony. 31 40
2. Hypocrisy is on the increase in our
society. 43 47 52
21. Even though we have reports in news-
papers, radio and television, it is hard
to get objective accounts of public
events. 49 29 39
29 Many major national sport contests
are fixed in one way or another. 38 49 30 50
18. Most elected public officials are really
sincere in their campaign promises. -39 48
6. Parents usually can be relied upon
ro keep their promises. -38 26
(Continued on next page)
 DIMENSIONALITY OF ROTTER INTERPERSONAL TRUST 1063

TABLE 1 (Cont'd)
CLUSTER
AND FACTOR
LOADINGS
OF M B ROTTER ~TERPERSONAL
TRUSTSCALBITEMS
Items Clusters Orthogonal Oblique Reducec
factors factors
Cl C 6 G VI V3 VS V I Bg Bs
32. Most salesmen are honest in describ
ing their products. -5 0 35 2
36. Most repairmen will not overcharge
even if they think you are ignorant
of their speciality. 4 1 59
5. Fear of social disgrace or punishment
rather than conscience prevents most
people from breaking the law.
9. The United Nations will never be an
effective force in keeping world
peace.
15. The future seems very promising.
23. Most experts can be relied upon to
tell the truth about the limits of
their knowledge.
24. Most parents can be relied upon to carry
out their threats of punishment.
39. Most people answer public opinion
polls honestly.
Proportion of variance accounted for 14 13 11 9 14 12 13
Note.-Decimal points omitted. Loadings < .30 omitted unless they occur for markers
of a given factor. Ttem numbers correspond to those of the 40-item Rorter Interpersonal
Trust Scale (25 test items plus 15 filler items). bItem-duster correlations, corrected for
item-duster overlap and cluster unreliability, are listed for the marker items of each cluster.
'Order in which items were added by the REDUCE routine.

first, second, and fourth in the amount of variance accounted for (accounting
altogether for 39% of the common variance). The correlations among these
three oblique factors were low (.23 between Factors 1 and 2, .27 between Fac-
tors 1 and 3, and -.06 between Factors 2 and 3). N o single oblique factor
appeared to match the fourth cluster. The "Oblique factors" column of Table 1
shows the loadings of the marker items for these three factors.
These findings suggest to us that there are four dimensions within the Rot-
ter Interpersonal Trust Scale for our college sample (although the fourth dirnen-
sion is common only to tlhe cluster analysis and orthogonal factor analysis).
The emergence of this dimensional structure is not attributable to the idiosyn-
crasies of any one analytic technique, since the techniques across which con-
vergence was found differed in terms of item weighting (unit vs fractional) and
relationship of dimensions (orthogonal vs correlated).
Examination of the loadings in Table 1 suggested a core of three marker
items for each of the four dimens~ons. These items were identified on the basis
of their consistent emergence across analytic techniques. Dimension I is defined
by Items 4, 13, and 40 (Irem 16 is not included because it fails to load on
1064 K-T. CHUN 8: J. B. CAMPBELL

orthogonal factor 1 and oblique factor 1 ) . These items focus on skepticism/

cynicism about politicians and political bodies (see Table 1 for actual items).
Cynicism appears to be the underlying theme, and we have labeled the dimen-
sion "Political Cynicism." Dimension 11, defined by Items 3, 14, and 26, con-
cerns self-protection or caution based on a perception of others as exploitative
and egocentric. W e have labeled it "Interpersonal Exploitation." Dimension
111, defined by Items 2, 21, and 29, concerns hypocrisy in our society and the
failure of impersonal, societal referents to meet commonly held expectations.
W e have labeled ir "Societal Hypocrisy." Dimension IV, defined by Items 6, 32,
and 36, also concerns hypocrisy and the failure to fulfill role requirements. How-
ever, here the referents are specific, personal roles, and the items are phrased in
a trustful direction. W e have designated this dimension as "Reliable Role-
perf~rmance."~
Although the need for sex-specific analyses has not been an issue with the
Rotter Interpersonal Trust Scale, it is always useful to establish the generalizabil-
ity of dimensional structure. Therefore, dimensional analyses were performed
on the subsample of 116 females using cluster analysis and factor analysis fol-
lowed by orthogonal rotation. (Similar analyses on the male subsample were
not attempted because of its small size.) Marker items for each dimension were
identified on the basis of consistent emergence across the two techniques and
then compared with those marker items identified for the total sample. There
was considerable correspondence between the resulcs from the total sample and
the female subsample on Dimensions I, 11, and 111. That is, all three marker
items appeared in the female subsample for Dimension 11, while two of the
three marker items emerged for Dimensions I and 111, i.e., Items 13 and 40 for
Dimension I, and 2 and 21 for Dimension 111. On Dimension N,however,
two marker items (32 and 36) appeared in the cluster analysis only, thus sug-
gesting that Dimension IV may be more salient for males.
The interrelationships among the four dimensions are quite interesting.
Both Dimensions I and 111 have impersonal referents such as "country," "in-
ternational politics," "sports contests," while Dimensions I1 and 1V refer respec-
tively to people in general and to people in specific social roles. Thus, only two
of the four dimensions are "interpersonal" in nature. Dimensions I and I11 can
be considered relevant to interpersonal trust and Rotter's definition thereof only
through an extension or modification of the definition. In addition to having
different referents, the four dimensions can be categorized by the nature of the
Person-Object (or Truster-Referent) relationship. Dimensions I, 111, and IV

'Kaplan (1973) reported three major factors underlying the Rotter scale for a sample of
97 college students. Judging from the exemplary marker items and the general descrip-
tion of individual factors, Kaplan's "Institutional Trust" and "Caution" factors appear
to resemble our "Political Cynicism" and "Interpersonal Exploitation" dimensions, re-
spectively.
 DIMENSIONALITY OF ROTTER INTERPERSONAL TRUST 1065

deal with hypocrisy and the fulfillment-nonfulfillment of role or normative

expectations; Dimension 11, however, concerns the attribution to people of such
characteristics as exploitativeness and selfishness. The "fulfillment of expecca-
tions" theme is obviously consonant with Rocter's definicion of interpersonal
trust, but the theme of "attributing 'Machiavellian' traits" is only secondarily
related. The attribution of such traits may conceivably lead to low interpersonal
trust, but this kind of causal chain is not a part of Rotter's definition. Thus the
link berween Dimension I1 and Rotcer's definition, if such a link exists at all,
can be established only by extrapolation.
These interpretations, which are based upon the item content of each
dimension, in turn suggese a particular pattern of incercorrelations. Dimensions
I and I11 should correlate highest, since they share the same referents and rela-
tionships. Dimensions I and IV, I11 and IV, and I1 and IV should have similar,
somewhat lower correlations, since these pairs share either the same referents or
the same relationship. Dimensions I and 11, and I1 and 111 should have the low-
est correlations, since they share neither the same referents nor the same rela-
tionship, i.e., 1-111 > I-IV, III-IV, II-IV > 1-11, 11-111. As shown in Table 2,
the correlations among the four dimensions do fit this pattern, i.e., .37 >
.29,
.29, .22 > .17, .15. The only exception is the slightly lower correlation between
Dimensions I1 and IV, but this lower value may suggest that relationship simi-
larity is more important than referent similaricy. (Assuming independence, the
over-all probability of these six values falling by chance in the predicted sequence
of inequalities is approximately ,001.)

TABLE 2
INTERCORRELATIONS AhD DESCRIPTIVE
STATIS~CS
FOR THE FOUR DIMENSIONS

Dimensions I I1 111 IV M SD
I .55" 6.97 2.37
I1 .17b*d .60 7.19 2.54
111 .37b-c .15cso .50 8.07 2.47
IV .2gd .22 .29" .50 8.05 2.20
"Diagonal contains coefficient alpha.
with identical superscripts are significantly different ( p
b-c~d."Coefficients < .O5, one-
tailed).

Table 2 also presents the mean score for each dimension, that is, the mean
of the sums of the three marker items for each dimension. A mean score of 9
would indicate neither trust nor distrust, with lower scores indicating lack of
trust. The mean scores for all four dimensions were significantly lower than
9 (p < ,001). A test for the significance of differences between dimension
means would be technically inappropriate, since the dimensions consist of dif-
ferent items whose units of measurement may differ. Noce, however, that the
mean scores for Dimensions I and I1 are considerably lower than those for
1066 K-T. CHUN 8r J. B. CAMPBELL

Dimensions I11 and IV, while the standard deviations are of comparable magni-
tude. The fact that the dimension which taps uust in politicians and political
bodies has the lowest mean score is fully commensurate wtih prevailing attitudes
in our society. It should be noted, however, that low trust in the political sphere,
i.e., Dimension I, does not logically preclude higher levels of trust in other
spheres, e.g., Dimensions I11 and IV. The Rotter Interpersonal Trust Scale,
however, fails to recognize that levels of trust can vary significantly across dif-
ferent sets of objects. This failure may lead to blurring of important differences.
The results from the application of the REDUCE routine indicate that the
Rotter Interpersonal Trust Scale is inefficient, i.e., it is possible to form a con-
siderably shortened version of the scale without substantive loss. The part-whole
correlation between the scale and its optimally constructed (in terms of maxi-
mizing the part-whole correlation) 12-item subset was .94. [The shrinkage in
cross-validation was .96 to .91 for the male-to-female, and .95 to .86 for the
femaleqo-male cross-validation (Chun & Pinneau, 1973, p. 4 4 ) .] The results
of the dimensional analyses and the REDUCE routine are quite consistent. Two
of the three marker items from each dimension are included in the optimal 12-
item subset. In fact, they represent eight of the first 10 items to be selected
(see the "REDUCE" column of Table I ) , thus providing further evidence to
support the proposed dimensional structure. The REDUCE algorithm con-
structs subsets which have maximum correlation with the total scale, and which
therefore tap all the variance in the whole scale. However, the 12-item subset
containing the marker items for the four dimensions taps only that portion of
the variance which is represented by these major dimensions. Consequently,
its part-whole correlation will be slightly lower than that for the REDUCE 12-
item subset (.89 vs .94). This slight disadvantage, however, is more than off-
set by the dimensional purity of the 12-marker item subset and by its higher
coefficient alpha (.69 vs .63 for the REDUCE 12-item subset). Although the
REDUCE 12-item subset is optimal as a short form of the Rotter Interpersonal
Trust Scale, we suggest that the subset comprised of the 12 marker items be used
when a shorter version is needed. This suggestion is based upon the clearer
dimensional structure and greater interpretability as a measure of trust of the
12-marker item set.
Mere demonstration of the factorial or dimensional structure of a measure
is inadequate: it is unsatisfactory for theoretical purposes because of the un-
known generalizabilit~,and it is insufficient for practical purposes since the
external validity remains to be determined. Dimensional analysis needs to be
followed by the demonstration of practical consequences, i.e., by the demonstra-
tion that conceptual clarification and increased predictability result from applica-
tion of the dimensionality. [For instructive examples of going beyond internal
analyses, see Krug and Moyer ( 1961), and Stein, Gough, and Sarbin ( 1966) .]
Unfortunately, despite our intention to carry the present examination beyond the
 DIMENSIONALITY OF ROTTER INTERPERSONAL TRUST 1067

restricted context of internal analyses, published data are not amenable to a

thorough reanalysis and raw data are not readily available. Consequently, we
content ourselves with a review of recent studies that lend credibility to our
dimensional structuring.
Evidence for the differential behavior of the four dimensions is provided
in the study by Hochreich and Rotter (1970). They present changes in mean
score for each of the 25 items over the 5-yr. period from 1964 to 1969. Not
a single marker item from our Dimension I1 (Interpersonal Exploitation) or
IV (Reliable Role-performance) is among those items that have shown signifi-
cant change, whereas all six items"comprising Dimensions I (Political Cynicism)
and 111 (Societal Hypocrisy) show significant change (see Table 2; Hochreich
& Rotter, 1970, p. 213). This clearly indicates that Rotter's measure is mulci-
dimensional, and that our dimensional strucruring is consistent with the actual
behavior of the items. These authors observe (p. 214) that the irems showing
significant change fall into two groups, one dealing with the establishment and
the ocher having to do with our society; this interpretation is clearly consistent
with our own, i.e., that Dimensions I and 111 represent Political Cynicism and
Societal H y p o c r i ~ y . ~
Since Wrightsman's "Trustworthiness" subscale in his Philosophies of
Human Nature Scale (Wrightsman, 1964) measures the degree of credibility,
dependability, or trustworthiness of people in general, we would predict a par-
ticular pattern of correlations between it and our four dimensions. Specifically,
the correlation of the Trustworthiness subscale should be higher with Dimensions
I1 and IV, which deal with personal referents, than with Dimensions I and 111,
which deal with impersonal referents. Since Wrightsman's subscale was a part
of the questionnaire administered to our sample, we had an opportunity to test
this prediction. The resulting pattern of correlations largely supported our
expectations. The Trustworthiness subscale correlated .52 with both Dimensions
I1 and IV, but only .32 with Dimension I and .47 with Dimension 111. The
difference between .52 and .32 is significant ( p < .01) for both the 1-11 and
I-IV comparisons; however, the difference between .52 and .47 is not signifi-
cant.
T h e item, '"The judiciary is a place where we can all get unbiased ueaunent," which is
included in our Dimension I, is not listed as one of the change items in Hochreich and
Rotter's Table 2. This appears to be a case of oversight or typographical error: the item
shows a mean change of -.3261 for females and -.2138 for males, while another item,
"Many national sports contests are fixed . . ." is listed as a significant change item wirh
the smaller mean changes of -.3222 for females and -.0229 for males.
'Hochreich and Rotter point oat that the irems exhibiting no significant change refer to
those objects with which one hns had some direct contact. From a broad social-psychologi-
cal perspective, a hypothesis has been proposed that the remoteness of personal contact with
the referent embedded in a test item increases that item's susceptibility to milieu effects
(K. T. Chun, Methodological explorations on the assessment of trust. Unpublished
manuscript, Univer. of Michigan, 1972). When the milieu is one of deepening or pre-
cipitating distrust, one would naturally expect, according to this hy thesis, that those
abstract items with relatively less personal contact would manifest a g r e a s e in trust.
1068 K-T. CHUN & J. B. CAMPBELL

There are several other reports in the literature in which scores on our
dimensions would probably have been better predictors than scores on the
Rotter scale as a whole. For example, consider Hamsher, Geller, and Rotter's
(1968) attempt to predict levels of belief or disbelief in the Warren Comrnis-
sion report on the assassination of President Kennedy, and Lotsof and Grot's
(1973) similar attempt for the Walker report on the Democratic Convention
disorders. W e would expect that Dimensions I (Political Cynicism) and 111
(Societal Hypocrisy) would have provided greater predictive power than the
total scale. Similarly, Geller (1967) probably would have done better using
Dimension IV (Reliable Role-performance), rather than the total scale, in his
attempts to predict Ss' trust in an experimenter, i.e., their reliance on E's assur-
ance of no further electrical malfunctions.
From the findings presented above, we conclude that the Rotter Inter-
personal Trust Scale is multidimensional for our college sample and thac this
multidimensional conception clarifies certain empirical findings of other scudies.
However, multidimensionality is not a part of Rotter's definition; thus there is
an incongruence becween his definition and measure of interpersonal trust.
W e now consider briefly the effects of response sets on the Rotter scale.
Little is known regarding the effects of social desirability other than the signifi-
cant correlations between the trusc scale and the Crowne-Marlowe Social Desir-
ability Scale (.21 for males, .38 for females, and .29 for males and females com-
bined; Rotter, 1967). These findings led Rotter to conclude that "trust is re-
garded as a socially desirable trait but thac the total amount of variance in the
trust scale accounted for by the social approval motive is relatively small" (1967,
p. 656). However, social desirability ratings of individual items are necessary
if the relationship between endorsement and perceived desirability is to be
assessed. For example, Rotter (1971) discounts the social desirability explana-
tion for the progressive decline in mean trusc score from 1964 to 1969 by noting
that the "drop is not the result of it becoming less socially desirable to express
trust" (p. 451). This conclusion is not warranted from his evidence thac the
"correlations with the Marlowe-Crowne Social Desirability Scale and the trust
scale did not differ significantly over the period studied" (p. 451). The evi-
dence which would permit Rotter's conclusion is the finding of no change in
the social desirability ratings of the items and the same degree of association be-
tween desirability rating and endorsement.
With regard to acquiescence, recall that the scale is balanced so chat agree-
ment wich 12 of the 25 items indicates trust and agreement with the remaining
13 items indicates distrust. Thus, the effects of acquiescence are controlled in
the usual manner. However, the acquiescence effect may take on a subtle form
in that it may be operative more in "agreeing with a trust item" than in "agree-
ing with a discrust item," or vice versa. The mean of the 12 trusc items (iM =
 DIMENSIONALlTY OF ROTTER INTERPERSONAL TRUST 1069

33.99, SD = 5.60), when prorated for 13 items (M = 36.83, S D = 6.07), is

indeed significantly greater rhan the mean of the 13 distrust items ( M = 32.81,
SD = 6.51; t = 8.92, p < .001). Of course, without partialling out the social
desirability-endorsement effects, this significant mean difference alone cannot
argue for the differential suengrhs of acquiescence in the trust and disuust
sets. Nevertheless, it suggests the possibility of differential effects and further
underscores the need for obtaining ratings on individual items.
The findings of the present study have several implications. First, we
recommend that scores on those dimensions identified in this report be used
along with the total Interpersonal Trust Scale score so that their incremental or
differential utility can be further established. Secondly, on a pragmatic level,
we also suggest that the set of 12 marker items may be used when a short
version of the Rotter Interpersonal Trust Scale is needed. Finally, it should be
noted chat the m a k e r items for the four dimensions could serve as four nuclei
for constructing expanded scales. The resulting dimensionally pure scales could
be used either individually or as components of a multidimensional measure of
trust.
REFERENCES
CHUN,K. T., & PINNBAU,
S. R.,JR. Improving testing efficiency through test abbrevia-
tion. Proceedings of the 81st Annual Convention of the American Psychological
Association, 1973, 8, 43-44.
CONSTANTINOPLE, A. An IEriksonian measure of personality development in college
students. Developme~talPsychology. 1969, 1 , 357-372.
CONSTAN-~INOPLB, A. Some correlates of average level of happiness among college
students. Developmental Psychology, 1370, 2, 447.
DBUTSCH, M. Trust and suspicion. Journal of Conflict Resolution, 1958, 2, 265-279.
ERIKSON, E. H. Childhood and society. New York: Norton, 1950.
ERIKSON, E. H. Identity: youth and crisis. New York: Norton, 1968.
GBLLER, J. D. Some persond and situational determinants of interpersonal trust. (Doc-
toral dissertation, Univer. of Connecticut) Ann Arbor, Mich.: Univer. Microfilms,
1967. (No. 68-1345)
G I P P ~K., The contribution of studies of source credibility to a theory of inter
rrust in the mmmunication process. Psyrhologica~Bulletin, 1967, 68, 10z2"O""'
HAMSHER, J. H., GELLW,J. D., & ROITER,J. B. Interpersonal n u t , internal-external
control, and the Warren Commission Report. Journal o f P e r s o d i t y and Social
Psychology, 1968, 9, 2 10-215.
HOCHREICH, D . J., & ROITER, J. B. Have college students become less trusting? lournal
o f Personality and Social Psychology, 1970, 15, 211-214.
KAPLAN,R. M. Components of trust: note on use of Rotter's scale. Psychological Re-
ports, 1973, 33, 13-14.
KRUG,R.E., & MOYER,K. E. An analysis of the F scale: 11. Relationship to standardized
personality inventories. Journal of Social Psychology, 1961, 53, 293-301.
LOTSOP,E. J., & GROT,J. S. Interpersonal trust, internal-external control and the Walker
Report on the Democratic Convention disorders. Psychological Reports, 1973,
32, 747-752.
ROITER,J. B. A new scale for the measurement of interpersonal trust. Jourml of
Personality, 1967, 35, 65 1-665.
ROITER,J. B. Generalized expectancies for interpersonal trust. American Psychologist,
1971. 26, 443-452.
 K-T.CHUN & J. B. CAMPBELL

SmGER, J. G., & SmGHR, D. G. Personality. Annual Review o f Psychology, 1972, 23,
375-412.
STEIN, K. B., GOUGH, H. G., & SARBIN,T. R. The dimensionality of the CPI socializa-
tion scale and an empirically derived typology among delinquent and nonde-
linquent boys. Multivariate Behavioral Research, 1966, 1 , 197-208.
WRIGHTSMAN, L. S . Measurement of philosophies of human nature. Psychological
Reports, 1964, 14, 743-751.

Accepted September 17, 1974.