Sei sulla pagina 1di 6

Journal oj Applied Psychology

1973, Vol. 60, No. 6, 736-741

Implicit Leadership Theory as a Determinant of the Factor


Structure Underlying Supervisory Behavior Scales
Dov Eden
Department of Labor Studies,
Tel Aviv University, Tel Aviv, Israel

Uri Leviatan
Center for Social Research on the Kibbutz,
Givat Haviva, Israel

Implicit leadership theory (preconceptions about the patterning of leadership


variables) was examined in a sample of 235 students. Students completed the
Survey of Organizations questionnaire on a fictitious "Plant X" about which
they were given little information. Factor analysis, performed on the items purported to measure four leadership factors, resulted in the conceptualized factor
structure. Since no information was given regarding supervisory behavior in
Plant X, the factor structure was attributed to an implicit leadership theory.
Factor analysis on subsamples indicated the factor structure could not be attributed to either experience in organizations or previous instruction in management. It is suggested that responses to questionnaires regarding organizational
variables may be contaminated by implicit theory and that multitrait-multimethod procedures may be required to validate questionnaires.

Several methodological investigations in the


area of personality assessment have addressed
the issue of whether the factor structure of
trait ratings reflects ratees' personality structure or merely prior conceptions concerning
personality traits, the so-called "implicit personality theory" that the raters bring to the
rating situation (Schneider, 1973). Typically,
persons with varying degrees of acquaintance
with the ratee are asked to describe him on a
set of trait scales. The resulting profiles are
factor analyzed, and factor structures for
raters with varying degrees of acquaintance
with the ratee are compared.
Earlier studies (Norman, 1963; Tupes &
Christal, Note 1, Note 2) found that similar
factor structures emerged when raters had
been acquainted with the ratees for as little
as 3 days or as long as 3 years. Mulaik
(1964) hypothesized that trait factors "correspond to distinct conventional concepts implied by the trait word" (p. S06). His results
showed that it is not necessary to rate actual
persons in order to obtain the personality
factors that would be associated with a set of
trait words. Using previously unacquainted
Yael Raveh and Sarah Arnon assisted. We thank
David G. Bowers, Allen I. Kraut, and Itai Zak for
comments on an earlier draft.
Requests for reprints should be sent to Dov Eden,
Department of Labor Studies, Tel Aviv University,
Ramat Aviv, Tel Aviv, Israel.

736

students' ratings of each other during the


first class meeting, Passini and Norman
(1966) obtained a factor structure highly
similar to that yielded by ratings of close
acquaintances and interpreted this as evidence
for implicit personality theory. Simply stated,
these studies have shown that a ratee is not
required to obtain remarkably stable and replicable personality factors.
Norman and Goldberg (1966) showed that,
notwithstanding the factor structure similarity, ratings obtained under implicit personality theory conditions differed in variance
from ratings of acquainted persons. An index
of "conspect reliability," which reflects the
degree of interrater agreement, was lower the
less the raters and ratees were acquainted and
was essentially zero for Monte Carlo data.
Furthermore, using peer ratings, self-ratings,
and predicted peer ratings, these researchers
computed a multitrait-multimethod matrix
and, on the basis of the convergent and discriminant relationships, concluded that more
prior acquaintance produces ratings with
greater relevance to ratees' personality characteristics.
Despite the reassurance implicit in Norman
and Goldberg's (1966) finding, they admitted
that their analyses controlled randomness
more convincingly than shared biases, stereotypes, or other systematic but trait-irrelevant
sources of variance, and the implicit per-

IMPLICIT LEADERSHIP THEORY DETERMINES FACTOR STRUCTURE


sonality theory continues to trouble investigators active in that area.
This issue has relevance for the study of
organizational behavior. Questionnaires filled
out by organizational members appear to be
the modal data collection method in the study
of organizational variables such as leadership.
If it can be shown that individuals respond to
such questionnaires in terms of implicit theory, many factors obtained through factor
analysis of the questionnaire responses may
be mere reflections of respondents' prior conceptions and not veridical representations of
empirical reality in the organizational environment. If the same pattern of item covariances obtained when members describe their
own organizations can also be obtained when
respondents complete the same questionnaire
under instructions to rate a fictitious organization, the conclusion that the factor structure was brought "in their heads" to the data
collection situation would be unavoidable.
This would render ambiguous the meaning of
previously obtained factor structures of organizational variables, since these structures
could have resulted from "real" organizational factors, or from respondents' conceptual factors, or from both. Such scales may be
reliable measures of (mis)conceptions rather
than perceptions. Organizational factors, like
beauty, may be in the eye of the beholder.
The present study was undertaken to seek
evidence for the existence of implicit theory
in rating scales commonly used in research
on organizations. Focusing on leadership
scales, we sought to replicate a factor structure obtained recurrently in past studies,
under conditions in which respondents could
not possibly be rating a supervisor's behavior.
METHOD
Sample
The respondents were 235 undergraduate and graduate students enrolled in courses in psychology,
sociology, labor studies, business administration, and
management during 1971-1973.

Questionnaire
The 1969 version of the Survey of Organizations,
which is reproduced in Taylor and Bowers (1970,
Appendix A-3), was translated into Hebrew by the

737

investigators for use in the present study. 1 The leadership scales were focused upon since they had been
included in the Survey of Organizations on the basis
of previous work by Bowers and Seashore (1966),
who conceptualized and measured four leadership
factors. These factors are (a) Supportbehavior that
enhances someone else's feelings of self-worth and
importance, (b) Interaction Facilitationbehavior
that encourages group members to develop close,
mutually satisfying relationships, (c) Goal Emphasisbehavior that stimulates an enthusiasm for meeting the group's goal or achieving excellent performance, and (d) Work Facilitationbehavior that helps
achieve work-related goals by such activities as
scheduling, coordinating, planning, and providing resources needed to get the job done.
These four factors were found by Halpin and
Winer (19S7) with the widely used Leadership Behavior Description Questionnaire developed at Ohio
State University and were labeled consideration, sensitivity, production emphasis, and initiating structure, respectively. Since sensitivity and production
emphasis accounted for little variance, they have been
largely ignored, leaving consideration and initiating
structure as the leadership factors commonly identified with the Ohio State University studies. Bowers
and Seashore nonetheless measured all four factors,
and they were all included in the present study due
to their demonstrated replicability.

Procedure
The questionnaire was administered during class
time. Students were instructed to use the scales in
the questionnaire to describe "a plant you do not
know. It is Plant X, it makes food products, and it
is located in the central region of the country." They
were told that this was being done for research purposes in order to learn more about the questionnaire
as a measuring instrument, and that prior experience
had shown that it was possible to fill it out under
such conditions. Those who complained of further
difficulty were instructed to "use your imagination."
Completion 'time was approximately 30 minutes.
After all questionnaires were collected, the respondents were briefed as to the purpose of the study
and their questionnaire experience was later used
pedagogically as an example of organizational measurement.

Analysis
Our aim was to replicate under implicit conditions
(i.e., the present instructions) the factor structure
obtained under normal conditions (e.g., "rate your
own organization"). Therefore, factor analyses were
performed to obtain principal components rotated
to a varimax solution. Squared multiple correlations
were used as communality estimates.
1
The authors are grateful to David G. Bowers for
his permission to translate and use the Survey of
Organizations for the present study.

Dov EDEN AND URI LEVIATAN

738

TABLE 1
FACTOR LOADINGS FOR FOUR LEADERSHIP FACTORS
Factor

1.

Leadership item"

Support

Friendly and easy to approach

67
74
82

2. Attentive to what you say


3. Willing to listen to your problems
4. Encourage effort
5. Maintain high standards
6. Example by hard work
7. Show you how to improve
8. Help you plan ahead
9. Offer ideas foi solving problems
10. Encourage team work
11. Encourage exchange of ideas
12. Frequency of group meetings

Work
facilitation

Interaction
facilitation

Goal
emphasis

54
58
41
72
72
71
61
82
56

.56
.62
.73
.43
.35

.36
.61
.53
.58
.62
.78
.40

Note. Only loadings of .40 or greater are shown. Decimal points are omitted from the loadings.
Item numbers correspond to those in Taylor and Bowers (1970, p. 75).

RESULTS
The factor analytic results for the 12 supervisory leadership items are shown in Table 1.
The factors, rotated to simple structure, explained 55% of the variance in the items and
replicated without exception the four factors
of leadership that the items were designed to
assess. The factors of Support, Work Facilitation, Interaction Facilitation, and Goal
Emphasis accounted for .34, .11, .06, and .04
of the variance, respectively. Using a .40 cutoff for loadings, each item loaded on one and
only one factor, in accordance with the fourfactor conceptualization. Indeed, the separation of the factors appears to be neater in
Table 1 than in Taylor and Bowers's (1970)
results obtained by performing a smallest
space analysis on responses to these same
items from employees in an oil refinery; their
results showed that several of the items did
not cluster empirically as conceptualized. We
know of no prior factor analytic validation of
these particular items purported to measure
the four factors of leadership postulated by
Bowers and Seashore (1966), though these
items were derived from items investigated in
previous factor analytic studies.
The results in Table 1 can be interpreted
as evidence that the four factors do indeed
represent conceptual factors that the respondents brought with them to the measurement

situation; that is, implicit leadership theory


determined the factor structure. Alternatively,
one could argue that the respondents were
given such an impoverished description of
Plant X that they responded according to an
image that they had of some organization
they were familiar with, such as a past or
present workplace, or according to preconceptions acquired in courses on management or
organizational psychology. The questionnaire
included questions about employment history
and a question asking what served as the
basis for choosing answers to most of the
items: (a) previous experience in a plant or
other type of organization, (b) experience on
a present job, (c) descriptions read about
work in plants, (d) random responding, or
(e) other. Additional factor analyses were run
on several subsamples defined by responses to
these questions.
The existence of implicit leadership theory
would be more convincingly confirmed if responses not influenced by actual experience
in work organizations yielded the factor structure. Inexperienced respondents would have
had less opportunity to make empirically
based observations of organizational phenomena and would have had to draw more heavily
on nonexperiential preconceptions about behavior in organizations. Only 81 respondents
admitted that, contrary to instructions, they

IMPLICIT LEADERSHIP THEORY DETERMINES FACTOR STRUCTURE


answered the questionnaire describing cither
a present or a previous workplace. The responses of the remaining 148 respondents,
who claimed that they answered on the basis
of descriptions they had read, or at random,
or "other," were factor analyzed separately,
For this subsample, the same four factors accounted for 58% of the variance in the items.
If respondents' self-reports about how they
answered are valid (and there is reason to
doubt their validity), the results for these
148 respondents indicate that the factor structure is not derived from empirical observation, thus making implicit leadership theory
an even more plausible explanation of the
findings.
A stronger test of the influence of experience on the factor structure could be made
using an organizationally naive sample. The
closest this could be approximated with the
present sample was to analyze separately the
responses' of those who had little or no work
experience as indicated by the employment
history questions. Naturally, all respondents
had experienced years of membership in educational and/or military organizations.
A rule of thumb in factor analysis is that
the ratio of respondents to items should be at
least 10 to 1 (Nunnally, 1967). For the 12
leadership items factor analyzed here, a minimum sample size should therefore be 120.
However, only a handful of respondents in
the present study had had no work experience. Combining those with three years of
experience or less yielded a sample of 99 relatively "inexperienced" respondents, whose responses were factor analyzed separately. The
remaining 119 respondents with 4 or more
years of work experience were also analyzed
separately.
The results for the inexperienced subsample were very similar to those in Table 1. The
only deviation from simple structure was a
loading of .46 for Item 12 on the Support
factor in addition to its loading on Interaction
Facilitation. The factor structure for the experienced subsample differed markedly from
the one in Table 1. The order of the factors
for this subsample was Work Facilitation, Interaction Facilitation, Support, and Goal Emphasis, accounting for .34, .12, .07, and .05

739

of the variance, respectively. Also, Item 1, a


Support item, loaded on Interaction Facilitation, and Item 6 did not load on any factor.
Thus, while the basic factor structure remained intact for both experience subsamples,
it was simpler and neater for those with less
work experience, despite the relatively small
number of cases. This further confirms implicit leadership theory. If the factor structure was derived from experience rather than
from preconceptions, it should get "cleaner"
with experience, not "messier."
Due to the widespread teaching of the Ohio
State University leadership studies in psychologically oriented management courses, it
is possible that our sample, composed predominantly of students, could be responding
in terms of these factors due to prior course
exposure to research results or to leadership
theory. Therefore, a factor analysis was performed separately on the responses of only
those 92 respondents who reported having
had no prior course in management or organizational science in any setting. Again, the
same basic four factors were obtained, explaining 51% of the variance in the items.
Two minor exceptions were that the loading
for Item 5 on Goal Emphasis dropped to only
.35, and Item 7 loaded .41 on Goal Emphasis
in addition to its loading on Work Facilitation. Thus, the factor structure shown in
Table 1 does not originate from course indoctrination.
The results of a separate factor analysis
for 46 respondents who claimed that they
filled out the questionnaire "completely at
random" showed that they did not respond
randomly at all. The five resulting factors left
Support and Work Facilitation intact. Item
4, a Goal Emphasis item (encourage effort),
separated from the rest and comprised the
fifth factor, its place being taken by Item 12
from the Interaction Facilitation factor.
Thus, the basic factor structure was largely
maintained even when only those who claimed
to have answered at random were included in
the analysis. Part of the deviation o f ' t h i s
structure from that obtained for the whole
sample undoubtedly derives from the very
small number of cases. Though smaller than
the commonly recommended sample size, the

740

Dov EDEN AND URI LEVIATAN

Some might argue that it makes little difference that respondents rate social objects,
real or fictitious, in terms of an underlying
conceptual structure. This conceptual structure derives from a lifetime of interaction
with people in organizational settings and as
such it reflects an underlying social reality.
DISCUSSION
We think this is circular reasoning. While the
Roach and Davis (1973) showed what they present results do not "prove" that convenconsidered "remarkable stability" in the fac- tional questionnaire assessment of organizator structure yielded by an attitude question- tional phenomena is invalid, they do show
naire administered twice in the same organi- that a leader is not required to obtain the
zation with a 10-year interval. They con- familiar leadership factors, thus raising the
cluded that psychometricians could therefore possibility that these factors are solely conbe more confident of the validity of such mea- ceptual. To say that the factors are valid besurement. The present study raises the ques- cause they derive from conceptions that are
tion of whether factor structures can be too learned from experience with social reality is
invariant. If a factor structure emerges even circular unless one uses an independent meaunder absurd instructions, this strengthens sure of social reality. Herein lies the remedy
the argument that it reflects something in the to the problem. Prior to Copernicus, it is unculture at large. Specific semantic character- likely that surveys about the nature of the
istics of the words cannot explain the invari- universe would have evidenced the heliocenance of the factor structure, since the present tric system. It is possible that after the
study was done using a Hebrew translation of Copernicus of leadership states his theory, the
the questionnaire.
present factors will be remembered as an
From what source do the factors in Table erroneous conception of an earlier age. It
1 originate? Certainly not from supervisory should be realized that questionnaires are
leadership behavior in Plant X. The respon- often purported to measure not merely peodents must have carried the patterned item ple's perception of and attitudes toward leadcovariances that produced the factor structure ership, but actual leadership behavior. We
into the data collection situation "in their think that measures well suited to assess attiheads." When the same factor structure is tudes are inappropriately used to measure the
obtained from respondents instructed to de- actual referents of those attitudes. This may
scribe their actual supervisors, does it reflect be the heart of the problem.
empirical reality, or respondents' preconcepThe problem is basically one of validity.
tions, or both?
Since questionnaire data may be manifesting
It would be interesting to find a sample or raters' conceptions only, the questionnaire
a set of instructions under which the factor should be validated against nonquestionnaire,
structure would not be obtained. Persons with observation-based measures of the same pheno experience in organizations could have nomena. The Campbell and Fiske (1959)
only the culture as a source for the factor multitrait-multimethod matrix is an approstructure. Organizationally naive youngsters priate paradigm. In addition, following Norwould make an interesting sample, but alas, man and Goldberg (1966), conspect reliabilby the time they have mastered the language ity could be used to differentiate less acwell enough to handle the questionnaire, they quainted raters and ratees from more achave already experienced membership in a quainted raters and ratees, since it is lower
formal organization, the school. Also, it would for the less acquainted persons. The more
be interesting to administer a set of leadership acquainted samples have been shown to proitems to persons with entirely different cul- duce ratings with greater relevance. Thus,
tures and values, such as career soldiers, indirectly, conspect reliability can be used to
Buddhist monks, Bedouin nomads, and others determine relevance by identifying better acto seek variations in the factor structure.
quainted raters.

results are mentioned because too small a


sample should detract from factor stability.
The factor structure is so robust that it was
obtained even from too small a sample of
purportedly random responses to a translation
of the original items.

IMPLICIT LEADERSHIP THEORY DETERMINES FACTOR STRUCTURE


In the meantime, it is evident from the
present results that the leadership factors are
in the mind of the respondent, It remains to
be established whether or not they are more
than that.
REFERENCE NOTES
1. Tupes, E. C., & Christal, R. E. Stability oj personality trail rating factors obtained wider diverse
conditions (USAF WADC Tech. Note 58-61).
Lackland Air Force Base, Texas: Personnel Laboratory, Air Force Systems Command, United
States Air Force, 1958.
2. Tupes, E. C., & Christal, R. E. Recurrent personality factors based on trait ratings (USAF ASD
Tech. Rep. 61-97). Lackland Air Force Base,
Texas: Personnel Laboratory, Air Force Systems
Command, United States Air Force, 1961.
REFERENCES
Bowers, D. G., & Seashore, S. E. Predicting organizational effectiveness with a four-factor theory of
leadership. Administrative Science Quarterly, 1966,
11, 238-263.
Campbell, D. T., & Fiske, D. W. Convergent and
discriminant validation by the multitrait-multimethod matrix. Psychological Bulletin, 1959, 56,
81-105.
Halpin, A. W., & Winer, J. A factorial study of the
Leadership Behavior Description Questionnaire. In
R. M. Stogdill & A. E. Coons (Eds.), Leader be-

741

havior: Its description and measurement (Research


Monograph 88). Columbus: Ohio State University,
Bureau of Business Research, 1957.
Mulaik, S. A. Are personality factors raters' conceptual factors? Journal of Consulting and Clinical
Psychology, 1964, 28, 506-511.
Norman, W. T. Toward an adequate taxonomy of
personality attributes: Replicated factor structure
in peer nomination personality ratings. Journal of
Abnormal and Social Psychology, 1963, 66, SV4583.
Norman, W. T., & Goldberg, L. R. Raters, ratees,
and randomness in personality structure. Journal
oj Personality and Social Psychology, 1966, 4,
681-691.
Nunnally, J. Psychometric theory. New York: McGraw-Hill: 1967.
Passini, F. T., & Norman, W. T. A universal conception of personality structure? Journal oj Personality and Social Psychology, 1966, 4, 44-49.
Roach, D. E., & Davis, R. R. Stability of the structure of employee attitudes: An empirical test of
factor invariance. Journal of Applied Psychology,
1973, 58, 181-185.
Schneider, D. J. Implicit personality theory: A review. Psychological Bulletin, 1973, 79, 294-309.
Taylor, J. C., & Bowers, D. G. The survey of organizations: Toward a machine-scored, standardized questionnaire instrument. Ann Arbor: University of Michigan, Institute for Social Research,
1970.
(Received August 28, 1974)

Potrebbero piacerti anche