Escolar Documentos
Profissional Documentos
Cultura Documentos
Religion?
Author(s): Lee A. Kirkpatrick and Ralph W. Hood, Jr.
Source: Journal for the Scientific Study of Religion, Vol. 29, No. 4 (Dec., 1990), pp. 442-462
Published by: Wiley on behalf of Society for the Scientific Study of Religion
Stable URL: http://www.jstor.org/stable/1387311 .
Accessed: 18/07/2014 13:57
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
Wiley and Society for the Scientific Study of Religion are collaborating with JSTOR to digitize, preserve and
extend access to Journal for the Scientific Study of Religion.
http://www.jstor.org
Although the Intrinsic-Extrinsic (I-E) frameworkhas provided the dominant conceptual and
measurementparadigmfor the psychology of religionover the last two decades,researchin this tradi-
tion suffers from a number of serious limitations. This paper details a variety of theoretical and
methodologicalcriticisms of contemporaryI-E research,many of which have been noted by previous
researchersbut have yet to be adequatelyaddressed.Theoreticalproblemsincludelack of conceptual
clarity in the definitions of I and E; confusion regarding what I and E are supposed to measure
(namely, intrinsic-extrinsicwhat?);the value-laden "good-religion-versus-bad-religion" distinction
underlyingthe framework;the problemsinherentin definingand studying religiousnessindependently
of beliefcontent; and the thornyissue of how I and E are conceptuallyinterrelated(namely,Allport's
originalbipolarcontinuumversus the moderntwo-factortheory). Criticismsof the measurementof
I and E concern the factorial structure, reliability, and construct validity of the I and E scales, as
well as the empiricalrelationshipbetween the scales. Although the I-E paradigm has clearly been
successful in terms of generatingresearchdata, it is our view that muchof this workis theoretically
impoverishedand has really taught us very little about the psychology of religion.Only a few years
after the Allport-RossI-E scales appearedon the scene, several researchersargued that it was time
to move beyonda simplistic I-E frameworkto more theoreticallyand psychometricallysophisticated
approaches.Nearly 20 years later, we believe this admonitionbears repeating.
INTRODUCTION
tLee A. Kirkpatrick was until recently a visiting assistant professor in the Department of Psychology, Univer-
sity of South Carolina, Columbia, SC. He is now a post-doctoral research associate in the Department of
Psychology, University of Massachusetts, Amherst, MA 01003, where all correspondence should be directed
Ralph W. Hood, Jr. is a professor of psychology at the University of Tennessee, Chattanooga, TN 37401.
? Journal for the Scientific Study of Religion, 1990, 29 (4): 442-462 442
Within four years of the publication of Allport's and Ross's (1967) work, however,
both the scales and the constructswereunderfire.Dittes (1971)and Hunt and King (1971)
suggested that the frameworkwas beginning to outlive its usefulness and called for con-
ceptually and psychometrically more sophisticated approaches to supersede it. Dittes
(1971:382)speculated, however, that the I-E typology showed "considerablepromise of
surviving [its] obituary."It now appearsthat he was right. Fourteenyears after Dittes's
remarks were published, Donahue (1985b:400)counted 70 published studies employing
Ailport's Religious OrientationScale and concludedthat "no approachto religiousness
has had greater impact on the empiricalpsychology of religion."
Manycriticismshave beenleveledagainst the I-E frameworkover the years by various
researchers;many of the important ones were pointed out long ago by Hunt and King
(1971)and by Hoge (1972).These old problems,it seems to us, have never been resolved
satisfactorily.In the meantime,new problemshave been identifiedthat also await resolu-
tion. We feel it is time to review these various criticisms, introduce some new ones, and
encourage researchers in the psychology of religion to pursue more promising
methodological and theoretical directions.
In his 1984AmericanPsychologist article,RichardGorsuchsuggested that measure-
ment has been both the boon and bane of the psychology of religion. This statement is
as trueof the I-E traditionin particularas it is of the measurement"paradigm"(as Gorsuch
described it) in general. There can be little doubt that this frameworkhas generated a
great deal of empiricalresearchin the psychologyof religionover the last coupleof decades,
and in this sense it has certainly been a boon to the field. However, we believe that in
many ways the I-E frameworkhas been a bane as well.
It is generally not a simple matter to disentangle conceptual, theoretical concerns
from empirical,psychometricones. For purposes of exposition, however,we will attempt
to do so to some extent, although the interconnectionswill become readily apparent(and
unavoidable)at various points.
reviewing Allport's various writings on the topic, Hunt and King (1971:340)concluded
that Allport'sdefinitionof I-E showed"a clearprogressiontowardviewingthe phenomena
as types of motivation,"that is, "themotives associatedwith religiousbeliefand practice."
Hoge (1972:370)agreed with this assessment, noting that I-E "is clearly a measure of
motivationfor religiousbehaviorratherthan the behavioritself.... We call this dimension
'intrinsicreligious motivation'. . . [and]use the term 'extrinsic motivation' for the latter
end of the dimension."
In the same paper, Hunt and King also suggested that in the work of Allport and
Ross (1967), as well as related work by Allen and Spilka (1976) concerning consensual
versus committed religion, I-E "began to look more like a personality variable and less
like 'religion'," adding that I and E "may prove to be basic pervasive personality
variables"(Hunt and King 1971:353;emphasis in original).With respect to the Extrinsic
scale in particular,they noted that with minorrewordingthe items might easily be applied
to any area of institutional behavior or involvement. Thus, one might be extrinsically
oriented toward almost anything, with religion representing only a special case.
McFarland's (1989a) adaptation of the I-E dimensions to Communist beliefs and com-
mitment in the Soviet Union offers a recent illustration of this interpretation. Kahoe
(1974:817)also seemed persuaded by the idea that an "underlyingpersonality variable
links [intrinsicreligion]and internal motivations." Allport and Ross (1967) themselves
discussed their results, particularly with respect to the Indiscriminate Proreligious
category, in terms of cognitive styles.
Examination of the item content of the Religious OrientationScale (ROS)corres-
pondingly reflects this conceptual confusion. Gorsuch (1984)has pointed out that these
items represent a "hodgepodge"of attitudes, beliefs, values, and behaviors. (The "Age
Universal" version of the I-E scales developed by Gorsuch and Venable [1983]partially
solves the problemby eliminatingbehavioritems.) The fact that the Intrinsic items, for
example, display relatively high intercorrelationsand tend to "hang together" in factor
analyses does not in itself justify aggregationof the items into a meaningfulscale. Under
the traditional common-factormodel, such aggregation implies that a latent construct
("intrinsicreligiousorientation,"whateverthat is) representsthe commoncause underlying
the inter-item correlations.' However, as Gorsuch (1984) has argued, the relationships
among values, attitudes, behaviors, and so forth constitute interesting empirical ques-
tions that should be resolved by research, not by fiat.
So, what does religiousorientationmean?Is it aboutmotivation,personality,cognitive
style, or something else? Perhaps it is a little of each; but in any case, greater precision
in definition is clearly called for from a scientific perspective. Sloppy definitions make
it difficult to devise and test precise hypotheses in empirical research. One particular
manifestationof this is the difficultyof assessing the construct validity of the scales used
to measure the constructs: How can we determinewhether the I-E scales are valid if the
constructs themselves are poorly defined?The credibilityand conceptualutility of a con-
struct derive from the logic of a theoretical system in which it is embedded.Conceptual
Putting aside for a moment the question of what religious orientationmeans, let us
consider the meaning of the more specific terms "intrinsic"and "extrinsic."Again, the
question of how I and E are definedor conceptualizedis somewhatdifferentfrom(though
not unrelatedto) the problemof how the psychometricinstrumentsused to measurethem
behave empirically.
In their comprehensivereviewof Allport'swritingson the topic, Hunt and King (1971)
called attention to a considerablelack of theoreticalclarity concerningAilport's multiple
usages of the terms "intrinsic"and "extrinsic."Specifically,they cited 11 distinct dimen-
sions or components used in Ailport's various writings to differentiate I and E. They
concludedthat the Intrinsic dimension is particularlyill-definedby this multiplicity of
diverse meanings, whereas the E dimension is more clearly defined as an "instrumen-
tal/selfish motivation"for religiousinvolvement and is adequatelymeasuredby Feagin's
(1964) six-item scale. Similarly, Dittes (1971:375)bemoanedthe "formaluntidiness" of
the I-E typology, which he characterizedas representing "multipledefining categories,
carelessly agglutinated, shifting from one discussion to another,not integratedwith each
other conceptually, and manifestly not correlated empirically."
This criticismof the I-E scale is relatedto the "hodgepodge"problemcited by Gorsuch
(1984)and is supportedby factor-analyticresults. Hunt and King (1971)cited their own
previouswork in which I and E items were factoranalyzedin the context of a wide variety
of other religion items. Although many of the E items loaded together on an extrinsic
motivation factor, the I items did not form a unitary factor but instead were scattered
across several different factors (King and Hunt 1969). This result suggests that the I
items are indeed tapping a diversity of dimensions of religiosity. We will have more to
say about these factor-analyticresults below.
Extrinsic-Social (Es) factor involving the use of religion as a means toward social gain,
and an Extrinsic-Personal (Ep) factor involving the use of religion as a means toward
gaining comfort, security, and protection. A second, equally significant finding was that
several E items loaded on neither of the extrinsic factors, but instead evinced strong
negative loadings on the I factor. These "residual"E items appear,in effect, to represent
negatively worded or reversed I items. McFarland(1989b)scored these items as a third
separate extrinsic scale which he labeled (afterreverse scoring) "religiousseriousness."
We will say more about these items below.
Similarfactor-analyticresults had been reportedpreviouslyby other researchers,but
for understandable reasons these results had generally gone unnoticed: Hoge (1972)
reported his results in a footnote, and Amon and Yela (1968)published their data in a
Spanish-languagejournal.The three-factorsolutionhas also been replicatedmorerecently
by Radikeret al. (1988),using the Feagin scale, and by Gorsuch and McPherson (1989)
using the Age Universal scale.2 However, Kirkpatrick(1989b) went a step further by
demonstratingthat these dimensionsrelateddifferentiallyto a numberof other variables.
For example,Ep correlatesmore highly with prayer,while Es correlatesmore highly with
churchattendance.Althoughhe did not successfullyreplicateKirkpatrick'sfactoranalyses,
McFarland(1989b)found the Ep and Es scales to correlatedifferentiallywith discrimina-
tion against various groups.
These results have a number of important implications for the present discussion.
First, the I items (alongwith a few reverse-scoredE items) tended to forma unitary factor
in these analyses, seemingly in contrast to the King and Hunt (1969)results reported
above. However, from a factor-analyticperspective there is no inconsistency here. In the
context of the I-E items alone, the I items correlate with each other more highly than
with E items and so form a unitary factor. In the context of a broaderdiversity of religion
items, the I scale breaks up into its various components, since each item can now "find"
similaritems with which to cluster. Thus a unitary intrinsicfactor is observedonly within
the relatively impoverished context of the E items.
A secondimplicationof these factor-analyticresults is that the Extrinsic scale appears
to tap morethan one type of extrinsic"motivation"for religiousinvolvement.This perhaps
should come as no surprise. If E is intended to represent an orientation involving the
use of religion as a means toward other ends, this clearly invites the follow-upquestion:
a means toward what other ends? The items contained in the Allport scales seem to tap
two such motives or categories of motives, namely social gain and personal comfort/pro-
tection. Interestingly,previous attempts have been made to expand I-E theoreticallyinto
a three-dimensionalframeworknot unlike the I-Ep-Es factor solution (Echemendiaand
Pargament 1982; Fleck 1981).
2. We should point out that other investigators have failed to replicate the three-factorsolution (McFarland
1989b;Radikeret al. 1989).The fact that the three-factorsolution is not highly reliableis not surprisinggiven
the fact that, dependingon the specific pool of I-E items used, the Ep and ES factors are defined by only two
or three items each. Therefore,randomfluctuations (dueto samplingerror)in the values of a few criticalcorrela-
tion coefficients powerfully influence the likelihoodof finding distinct Ep and ES factors.
If we grant that I does represent some sort of unitary dimension, what exactly is
this dimension?Batson (1976)and Batson and Ventis (1982)have claimedthat it essen-
tially measures a religiousorientationreminiscentof Hoffer's(1951)"truebeliever,"that
is, a "singlemindedness"or rigid adherenceto doctrinalorthodoxy. Batson proposedthe
dimensionof religionas quest to representan aspect of maturereligionwhich, in his view,
was an importantcomponentof Allport's conceptualizationof maturereligionnot tapped
by the Intrinsic scale.3While Donahue (1985b)has arguedthat the "workshopof doubt"
idea is less centralto Ailport'sconceptualizationof maturereligionthan Batson suggests,
Hood (1985)has maintainedthat overcomingdoubt is central to Ailport's conceptualiza-
tion of intrinsic religion. This debate is particularlyinteresting in the present context,
for it points once againto the diversity of theoreticalinterpretationsto whichthe Intrinsic
construct can be (and has been) subjected.
Anotherhint as to what the Intrinsicscale "really"measuresis providedby examina-
tion of the scale's empiricalcorrelates.Gorsuchand McFarland(1972)have demonstrated
that a single item ("Howimportant is your religion to you?")correlateswith the I scale
about as strongly as respective reliabilities will allow. Donahue (1985b)cites four addi-
tional studies in which a meta-analysis revealed a correlationof .76 between I and self-
rated importance of religion or religious commitment. These findings suggest that,
whatever its other psychometric and theoretical properties,the Intrinsic scale behaves
empirically as a measure of "religious commitment" (Donahue 1985b).
We interpret these data as pointing to a weakness of the I-E scales, not a strength.
There is no doubt that strength of religious commitment is an important variable to be
consideredin psychology of religionresearch,but it is a theoretically impoverishedone.
With this conceptualizationwe have wandereda long way from religious orientation as
a personality, cognitive, or motivational construct, and even further from Ailport's rich
theoretical descriptions of mature-immatureand intrinsic-extrinsicreligion. Moreover,
examinationof the scale'sitem content suggests that severalitems areprobablytangential
to the "commitment"interpretation,and that the scale should be modifiedsubstantially
if "religious commitment" is indeed what we intend for it to measure.
What is measuredby the Extrinsic scale? Hunt and King (1971)concludedthat the
scale measuresa selfish/utilitarianmotivation for religiousinvolvement:Donahue(1985b:
416) concludedthat E "does a good job of measuringthe sort of religionthat gives religion
a bad name." As described above, Kirkpatrick's(1989b)analyses suggest the presence
of two distinct extrinsic factors, although these are perhaps sufficiently intercorrelated
to justify aggregating them into a single scale for some purposes. (However, as noted
3. A critique of Batson's work is beyond the scope of the present article. Although his three-dimensional
frameworkof InternalEnd,External/Means,and Quest historicallyemergedout of the Ailporttradition,a review
and criticalanalysisof this workwouldtake us far afieldfromthe particularargumentsandissues beingdiscussed
here.
earlier,these analyses also show that several "residual"E items clearly do not belong
on the same scale with the Ep and Es items.)
It is important to note that the two extrinsic factors (Ep and Es) described by
Kirkpatrickwere decidedlynot theoreticallyderived.Two distinct E factorswerecertainly
not intended by Ailport (norwere any two factors, for that matter). The appearanceof
"extra"factors shouldproperlybe interpretedas a psychometricdeficiencyof the scales,
not as a substantive researchfinding. It thereforedoes not necessarily follow from these
results that researchersshouldnow begin to focus on, or developexpandedand "improved"
scales for measuring,extrinsic-personaland extrinsic-socialreligiousorientations.Gorsuch
(1988:212)has referredto the I/Ep/Es distinction as the "cutting edge in the utilization
of [the I-E] scales" and has called for improved measurement and additional research
on the Ep and Es dimensions (see also Gorsuch and McPherson 1989). In contrast, we
maintain that a serious approach to researching the topic of motivation for religious
involvement, if motivation is indeed the construct of interest, should begin with a rich
psychological theory of human motivation, not with a psychometric blunder.
Summary
A fundamentalstumbling block for the I-E frameworkfrom the very start has been
the question:Do I and E representopposite poles of a single continuum,or two indepen-
dent dimensions?Allport's theory clearly specified bipolaropposites, but the two-factor
theory has predominatedsince the early empiricalwork of Ailport and Ross (1967)and
Feagin (1964).Hoge (1972:370)clearly struck at the heart of the matter when he wrote:
"Theresearchertoday is faced with the question - is the weakness of the existing scales
due to Allport'stheoreticalimprecision,or is it due to poorscale constructionwhichallowed
'uninvited' factors to enter into the measurements?There is evidence supporting both
contentions." Hoge's remarksare no less applicabletoday than they were 18 years ago.
The problem has never really been solved.
Although we do not necessarily wish to be apologists for Ailport's earlierposition
regardingI and E as a single bipolardimension,we do believe that the one-factortheory
was abandonedmuch too easily and has never really received a fair empiricaltest. As
noted by Hoge (1972),the lack of fit between theory (onedimension)and data (two dimen-
sions)in Feagin's(1964)factor-analyticresults couldhave beenresolvedeitherby modifying
the theory to accommodatethe empiricalfactor structure, or by modifyingthe item pool
to produce a factor structure more congruent with the theory. Feagin concluded, and
Allport and Ross (1967) concurred,that the evidence proved I and E to be orthogonal
rather than bipolar. Of course, theories should be modified by data, but in this case it
seems that a long history of theoreticalworkwas abandonedon the basis of a single ques-
tionablefactor analysis and an exploratory,untested sample of questionnaireitems.4 As
one of us has argued elsewhere (Kirkpatrick1989b), this seems like an instance of the
tail wagging the dog. Given the rich theoretical work of Ailport, was it really sensible
to abandonthe bipolarconceptualizationso quickly?Few researchersseem to have con-
sideredthe alternative and at least equally plausible conclusionthat the initial I-E item
pool was simply inadequateto assess Allport's constructs. Only Hoge (1972)has taken
up the gauntlet and attempted to create a unidimensionalscale of "intrinsic religious
motivation," focusing on the distinction between ultimate (intrinsic)and instrumental
(extrinsic)religious motivation.
Conceptually,the notion of orthogonal I and E dimensions seems problematic.Of
course, this depends on the precise meanings assigned to I and E, a thorny issue which
has been discussed in some detail above. But if I and E are defined,for example, in terms
of the instrumental-ultimatedistinction used by Hoge (1972), they seem clearly to be
opposites. Can a person hold religion as a primaryvalue and master motive, and at the
same time merely be using religionas a means to other selfish ends? Again, a rich theory
of motivation is required to clarify the issue. For instance, ambivalence is central to
psychodynamicmotivational theories, thus allowing for "mixed" motivations.
Indiscriminate Proreligiousness
In any event, Allport and Ross (1967) discovered that a sizeable number of their
subjects persisted in endorsing both I and E items, and thus proposed a category of
IndiscriminateProreligiousness(IPR) to which these subjects were assigned - in con-
trast to pureIntrinsics(high I, low E) and pureExtrinsics (lowI, high E). Donahue(1985b)
found an overall I vs. E correlationof -.20 in his meta-analysisof existing I-E research.
4. We refer to Feagin's (1964) factor analysis as "questionable"because his methods and results are not
presentedin sufficientdetailto permita carefulcriticalevaluation.For example,the fact that his first two factors
accountedfor only 29% of the total variance suggests that extraction of additionalfactors might have been
warranted,althoughit is difficult to interpret"29%of the variance"without additionaldetails of his procedures.
The small magnitude of this correlationimplies, as observed by Ailport and Ross, that
a substantial numberof subjects do endorse both I and E items, but what exactly does
this pattern of indiscriminateproreligiousnessmean?Several alternativeinterpretations
are possible, but each has important shortcomings.
First, the IPR category (which might be operationally defined in any number of
ways - see below) might be interpreted as simply a third type of religious orientation
towardreligion,in much the same way as androgynyhas been defined as a third psycho-
logical sex-roleorientationalong with masculinity and femininity(Spenceand Helmreich
1978). The problemwith this interpretationis that, as we have seen, this third orienta-
tion is not easily accommodatedtheoretically. In contrast, androgyny (scoring high on
both masculinity and femininity)is not self-contradictorybecause the two dimensions
were conceptualizedand deliberatelyoperationalizedas orthogonal rather than bipolar
constructs. Our conceptualizationof the meaning of I and E orientations would need to
be modifiedconsiderablyto make room for a third IPR category as a meaningful,alter-
native orientation toward religion.
Ailport himself opted for a second alternative.Rather than cast IPR as a third orien-
tation towardreligionper se, Allportand Ross (1967)interpretedthe IPR responsepattern
as a response set bias or cognitive style somewhat akin to Rokeach's (1960) construct
of dogmatism. Their reasoning seemed to suggest that the IPR category represents a
qualitatively differentkind of variablefrom that reflected in the I and E categories. Un-
fortunately,however,this makes for a ratheruntidy conceptual/psychometricframework.
Among people who are classified as Intrinsics or Extrinsics, we are measuring an orien-
tation toward religion;for people who are classified as IPR, however, we are measuring
"cognitive style." Logically this seems roughly analogous to classifying people into the
three categories of Introverts, Extroverts, and Auto Mechanics. The confounding of
cognitive style or response set with the I-E constructs themselves adds further concep-
tual confusion to an already clouded picture (see McCallister1988 for a treatment of I,
E, and IPR in cognitive terms).
Pargamentet al. (1987)have taken a relatedapproachin suggesting that indiscriminate
proreligiousnessis a usefulconstructin its own right and have proposeda scale formeasur-
ing it directlyratherthan as a by-productof the I-E scales. Theirscale appearsto measure,
in effect, a social desirabilityresponse set specific to religionand thereforeseems concep-
tually distinct fromthe Intrinsic-Extrinsicframework.Pargamentet al. (1987)recognized
this in noting, "It is surprising... that efforts to deal with this problemhave beenlargely
limited to work in the specific area of intrinsic and extrinsic religious orientation."We
agree that the assessment of response-set biases or artifacts in religion measures is an
importantproblem,and that the approachsuggested by Pargament et al. may prove to
be a useful one. However,their conceptualizationand operationalizationof indiscriminate
proreligiousnessseem sufficientlydistinct fromthe I-E frameworkper se that it probably
warrantsa distinct label.This is theoreticallya more elegant solutionthan Ailport's,since
it defines and measures the response-set or cognitive style variable independently of
religious orientation rather than confounding the two constructs. Nevertheless, this
solution does not solve the problemof what to do with those "muddleheads"who score
high on both I and E.
This problemof what to do with subjects who endorse both I and E items has never
been solved satisfactorily.The most popularsolutioninvolves dichotomizingthe two scales
into "low" and "high" categories and then creating the well-knownfourfold typology.
However,this solutionsuffersfroma numberof interrelatedconceptualand methodological
problems.
First among these concerns is the fundamentalissue of "types" or class variables
versus continuous dimensions or dimensional variables. In most areas of contemporary
psychology,and particularlypersonalityresearch,individualdifferencesare conceptualized
almost exclusively in terms of dimensions rather than types. Discrete categories (for
example, introverts and extroverts) are sometimes employed as a matter of convenience
to illustrateopposingpoles of a continuum,but ultimatelythe underlying(latent)variables
are typically conceived as continuous dimensions. This perspective fits naturally with
most commonsensenotions of individualdifferences:We think of peopleas "moreor less"
(that is, as a matter of degree) fastidious, intelligent, warm, and so forth. Although we
might describe someone as a shy person, as if to imply that people are either shy or not
shy, what we usually mean is that the person falls toward the shy pole of a continuum.
With some reflectionwe couldprobablythink of other individualswho are somewhatmore
shy, or slightly less shy, than this person. In psychologicalresearch,this conceptualiza-
tion is reflectedin the 5-, 7-, or n-point responsescales whichcomprisemost psychometric
instruments.
Under certain conditions,however,reasonableargumentscan be made that a (latent)
variable is "truly" discrete. Defending such a position empiricallyis rather a difficult
and complexmatter rarelyattemptedin personalityresearch.(See Gangestad and Snyder
1985 for an exceptionand a detailedanalysis of these issues.)A seriousdefenseof religious
orientation as a "true" class variable, however, has yet to be undertaken.There is no
obvious empiricalevidence to support I and E as types: For example, the Intrinsic and
Extrinsic scales do not tend to producethe kind of bimodalfrequencydistributions one
would expect if there really were two types being imperfectly measured by continuous
scales. Again, this reflects our commonsensenotion that people vary along a continuum
with respect to their level or degree of personal commitment to religion, or the extent
to which they rely on religion for personal or social rewards.
We suspect instead that most I-E researchers,like other personality researchers,
generally utilize types merely as a matter of convenience, usually in order to employ
analysis of variance (ANOVA)for assessing possible statistical interactions between I
and E. However, this approachsuffers from two important drawbacks.First, collapsing
a continuous variable into a dichotomy discards a tremendous amount of information.
Arbitrary dichotomization of a continuous variable results in a considerable loss of
statistical power (Cohen1983 as cited in Pargament et al. 1987). Second, dichotomizing
the I and E scales precludesthe possibility of assessing curvilinearrelationshipsbetween
I and E and other variables (Pargament et al. 1987).
Another majorproblem,which is simultaneously theoretical and methodological,is
that there is no obvious basis for choosing cutpoints. The commonly used median-split
approachis problematicfrom a methodologicalstandpoint because the variation in cut-
Kirkpatrick's(1989b)analyses of the I-E scales shed some new light on the I-E inter-
correlationand the issue of indiscriminateproreligiousness.These results revealed that
(1) the Extrinsic-Personaland Extrinsic-Socialitems/scales are uncorrelatedwith I, but
(2) several of the "residual"items correlatestrongly and negatively with I and load with
I items in factoranalyses.It is these latter items whose contentindeedappearsto represent
the negationof I items. Examplesof such items include"AlthoughI am a religiousperson,
I refuse to let religious considerationsinfluencemy everyday affairs" and "Although I
believe in my religion, I feel there are many more important things in my life." These
data demonstrate that the negative correlationbetween I and E is due entirely to the
misplacementof these reversedI items on the E scale - that is, a psychometricdeficiency.
If these items were moved from the E scale to the I scale and reverse-scored,I and E
would be orthogonal rather than negatively correlated.
These residual E items were undoubtedly the ones Ailport had in mind when he
referredto the indiscriminatelyproreligiousas "muddleheads":People endorsing these
META-THEORETICALISSUES
Another problemwith I-E, in our view, is that it is part of a long tradition of distinc-
tions between "good religion"and "bad religion."In the words of Dittes (1971:375),the
I-E distinction carries "a heavy contrabandload of value judgment that simply will not
be sloughed off." Dittes addedthat this weakness of the I-E typology (as well as the lack
of theoreticalclaritymentionedabove)exists becauseits developershave been "principally
concernedabout the purity of religion, not about the purity of concept" (375). He refers
to Ailport's approachas "prophetic,"by which he means "being animated principally
REFERENCES
Allen, Russell 0. and Bernard Spilka Allport, Gordon W. and J. Michael Ross
1967 Committed and consensual religion: A 1967 Personal religious orientation and pre-
specification of religion-prejudice relation- judice. Journal of Personality and Social
ships. Journal for the Scientific Study of Psychology 5(4):432-443.
Religion 6(2):191-206. Amon, Jesus and Mariano Yela
Ailport, Gordon W. 1968 Dimensiones de la religiosidad. Reuista de
1950 The individual and his religion: A psycho- Psicologia General y Aplicada 23:989-993.
logical interpretation. New York: Batson, C. Daniel
Macmillan. 1976 Religion as prosocial: Agent or double
1954 The nature of prejudice. Cambridge: agent? Journal for the Scientific Study of
Addison-Wesley. Religion 15(1):29-45.
Batson, C. Daniel and W. Larry Ventis scale. Journal for the Scientific Study of
1982 The religious experience: A social-psycho- Religion 22(2):181-187.
logical perspective. New York: Oxford Griffin, Glenn A. Elmer, Richard L. Gorsuch, and
University Press. Andrea-Lee Davis
Benson, Peter L. 1987 A cross-cultural investigation of religion
1989 Defining and measuring religious orientation, social norms, and prejudice.
typologies. Paper presented at the conven- Journal for the Scientific Study of Religion
tion of the American Psychological Assoc- 26(3):358-365.
iation, New Orleans, August. Herek, Gregory M.
Dittes, James E. 1987 Religious orientation and prejudice: A com-
1971 Typing the typologies: Some parallels in the parison of racial and sexual attitudes. Per-
career of church-sect and extrinsic-intrinsic. sonality and Social Psychology Bulletin
Journal for the Scientific Study of Religion 13(1):34-44.
10(4):375-383. Hoffer, Eric
Donahue, Michael J. 1951 The true believer. New York: Harper.
1985a Intrinsic and extrinsic religiousness: The Hoge, Dean R.
empirical research. Journal for the Scientific 1972 A validated Intrinsic Religious Motivation
Study of Religion 24(4):418-423. Scale. Journal for the Scientific Study of
1985b Intrinsic and extrinsic religiousness: Religion 11(4):369-376.
Review and meta-analysis. Journal of Per- Hood, Ralph W., Jr.
sonality and Social Psychology 48(2): 1978 The usefulness of the indiscriminately pro
400-419. and anti categories of religious orientation.
Echemendia, Ruben J. and Kenneth I. Pargament Journal for the Scientific Study of Religion
1982 The psychosocial functions of religion: 17(4):419-431.
Reconceptualization and measurement. 1983 Social psychology and fundamentalism. In
Paper presented at the convention of the Rural psychology, edited by A. W. Childs
American Psychological Association, and G. B. Melton, 169-198. New York:
Washington, DC, August. Plenum.
Feagin, Joe R. 1985 The conceptualization of religious purity in
1964 Prejudice and religious types: A focused Ailport's typology. Journal for the Scien-
study of Southern fundamentalists. Journal tific Study of Religion 24(4):413-417.
for the Scientific Study of Religion 4(1): Hood, Ralph W., Jr. and Lee A. Kirkpatrick
3-13. 1989 Religious fundamentalism conceptualized
Fleck, J. Roland in terms of Rokeach's dogmatisnm, psycho-
1981 Dimensions of personal religion: A analytic and attachment theories. Paper
trichotomous view. In Psychology and presented at the meeting of the Society for
Christianity, edited by J. R. Fleck and the Scientific Study of Religion, Salt Lake
J. D. Carter, 66-80. Nashville: Abingdon. City, October.
Gangestad, Steve and Mark Snyder Hunt, Richard A. and Morton King
1985 "To carve nature at its joints": On the ex- 1971 The intrinsic-extrinsic concept: A review
istence of discrete classes in personality. and evaluation. Journal for the Scientific
Psychological Review 92(3):317-349. Study of Religion 10(4):339-356.
Gorsuch, Richard L. James, William
1984 Measurement: The boon and bane of in- 1902 The varieties of religious experience. New
vestigating religion. American Psych- York: Longmans, Green.
ologist 39(3):228-236. Kahoe, Richard D.
1988 Psychology of religion. Annual Review of 1974 Personality and achievement correlates of
Psychology 39:201-221. intrinsic and extrinsic religious orienta-
Gorsuch, Richard L. and Sam G. McFarland tions. Journal of Personality and Social
1972 Single vs. multiple-item scales for measur- Psychology 29(6):812-818.
ing religious values. Journal for the Scien- Kennedy, K. and Richard L. Gorsuch
tific Study of Religion 11(1):53-64. 1989 Normative data on intrinsic and extrinsic
Gorsuch, Richard L. and Susan E. McPherson religiousness. Paper presented at the
1989 Intrinsic/Extrinsic measurement: I/E- meeting of the Society for the Scientific
Revised and single-item scales. Journal for Study of Religion, Salt Lake City, October.
the Scientific Study of Religion 28(3): King, Morton B. and Richard A Hunt
348-354. 1969 Measuring the religious variable: Amended
Gorsuch, Richard L. and G. Daniel Venable findings. Journal for the Scientific Study of
1983 Development of an "age universal" I-E Religion 8(2):321-323.