Donahue JPSP 1985 Religiosidad Metanálisis

Embed Size (px)

Citation preview

  • 7/25/2019 Donahue JPSP 1985 Religiosidad Metanlisis

    1/20

    Journal

    of Personality and

    SocialPsychology

    1985,

    Vol.

    48, No. 2,

    400-419

    Copyright

    1985by the

    American

    Psychological Association, Inc.

    0022-35I4/85/S00.75

    IntrinsicandExtrinsic Religiousness: Reviewand Meta-Analysis

    MichaelJ. Donahue

    Brigham

    Young

    University

    Themajorfindingsof thismeta-analyticreview

    concerning intrinsic

    andextrinsic

    religiousnessarethese:(a)Samplesconsisting ofrespondentswithconservative

    theological

    orientationsseem

    morelikelyto

    display

    a

    negative correlation between

    intrinsic and

    extrinsic

    religiousness

    than

    do

    others,

    (b) Extrinsic religiousness

    tends to be

    positively

    correlated

    with

    negatively evaluated characteristics, and

    uncorrelated with

    measures

    of religious belief and commitment, (c) Intrinsic

    religiousness tends to beuncorrelated

    with

    negatively evaluated characteristics,

    and positively

    correlatedwithmeasures

    of

    religiousness,

    (d) A

    fourfold

    typology

    based

    on median splits of the two

    scales

    is of

    little

    use when the dependent

    variable is

    religious

    innature,but

    with

    various nonreligious variables produces

    results that may

    correspond

    to findings of

    curvilinearity

    observed

    with

    other

    measures

    of

    religiousness.

    Recommendationsconcerningthe use of theintrinsic

    and

    extrinsicscales in

    future

    research are made. The article concludes

    with

    a

    review

    ofrecent conceptual

    developments

    byBatson

    (1976)

    and

    Hood

    (1978).

    No approach

    to

    religiousness

    has had

    greater impact on the

    empirical

    psychology

    ofreligion than Gordon W.Allport's

    concepts

    ofintrinsic (7)

    and

    extrinsic(E) religiousness

    (Meadow & Kahoe, 1984).'Nearly 70 pub-

    lished studies have used Allport's Religious

    Orientation Scale (ROS), making it one of

    the most frequently used measures of reli-

    giousness.

    Research concerning / andE has been

    reviewed in three current

    psychology-of-reli-

    gion textbooks (Batson & Ventis, 1982;

    Meadow &Kahoe, 1984;Paloutzian, 1983).

    Although

    thesereviewsare

    useful,

    they have

    not closely examined

    a

    number

    of

    relevant

    issues,

    such

    as the

    I-E

    correlation

    and the

    I-E interaction. In addition to addressing

    these issues, in the present review I seek to

    applythetechniquesof meta-analysis(Glass,

    This article is a revision andexpansion of areview

    that originally appeared

    in the

    author's

    doctoral

    disser-

    tation, submitted

    at

    Purdue University.

    It was

    alsopre-

    sented

    at the

    meeting

    of the

    American Psychological

    Association, Anaheim,

    California,

    August 1983.

    Thanks are due to Richard Gorsuch, Allen Bergin,

    Richard

    Williams,JosephRychlak,

    Alice

    Eagly,

    Raymond

    Paloutzian, Peter Benson,

    Kay

    Deaux.Richard Heslin,

    HarieyBernbach, and an anonymous reviewer fortheir

    commentsonearlier drafts.

    Requests for reprints should be sent to Michael Don-

    ahue, who is now at

    Search Institute,

    122

    WestFranklin,

    Suite

    215, Minneapolis, Minnesota 55404.

    McGaw,

    &Smith,1981;Hunter, Schmidt,&

    Jackson, 1982)

    toI-E

    research

    in an

    attempt

    to explain a number of seemingly inconsistent

    fin ings

    EarlyConceptual Development

    Hunt

    and

    King (1971)

    reviewed the

    early

    conceptual history of 7 and

    E.

    In brief,

    Allport

    distinguished between

    two

    types

    of

    religious sentiment: Intrinsic religiousness is

    religion

    as a meaning-endowing framework

    in

    terms

    of

    which

    all of

    life

    isunderstood;it

    is religion

    as

    proto-point (Rychlak, 1977).

    Extrinsic religiousness,

    in contrast, is the

    religion of comfort and

    social

    convention, a

    self-serving,

    instrumental approach shaped to

    suitoneself. Table

    1 is a

    summary

    ofAllport's

    writings on these

    points.

    2

    7 and

    E

    were

    probably best summed up by Allport and

    Ross (1967) when they stated that "the

    ex-

    1

    Throughout thisreview,thetermreligiosityisavoided.

    The reason

    is

    etymological:

    "Religiosity"

    connotes

    an

    affected,

    artificial, or

    exaggerated religious interest; "re-

    ligiousness" doesnotcarry that

    conceptual

    baggageand

    isthereforemore appropriate in thepresent context.

    2

    Table 1 is similar to a table presented by

    Hunt

    and

    King

    (1971).

    However, Hunt

    and

    King examined

    /and

    Ein a predominantlysociologicallight and omitted such

    characteristics as maturity

    from

    their analysis. They also

    omitted Allport's (1961, 1963, 1966a)references,which

    stressthe

    mental health

    aspectsof theconcepts.

  • 7/25/2019 Donahue JPSP 1985 Religiosidad Metanlisis

    2/20

    I-E META-ANALYSIS

    401

    Table

    1

    Concepts Associated with Intrinsic and Extrinsic

    Religiousness

    in Allport's Writings

    Intrinsic

    Extrinsic

    Relatesto all of life

    (a,

    b, c, d, f, g, h, j)

    Unprejudiced;tolerance

    (a,

    b, c, h, i)

    Mature(a, d)

    Integrative;

    unifying;

    meaning-endowing

    (a, c, d, f, g, h, i)

    Regularchurchattendance

    (e, g, h)

    Makes

    for

    mental

    health

    f, g)

    Compartmentalized

    (a, c, d, h)

    Prejudiced;exclusionary

    (a,

    b,

    c, d, e, h)

    Immature;dependent;

    comfort;

    security

    (a, b, d, f,g,

    h,

    i, j)

    Instrumental; utilitarian;

    self-serving

    (a, c, d, e,

    f,g,h,i,j)

    Irregular church

    attendance(e,g,h, i)

    Defense

    orescape

    mechanism(d, f, g)

    Note.

    Lettersinparentheses

    refer

    to thefollowing

    refer-

    ences:

    (a)Allport (1950), (b)

    AUport (1954),

    (c) Allport

    (1959), (d)Allport

    (1961),

    (e)Allport

    (1962),

    (f)Allport

    (1963),(g) Allport

    (1964),

    (h)Allport

    (1966a),

    (i)Allport

    (1966b),

    (j) Allport &

    Ross

    (1967).

    trinsicallymotivated person uses

    his

    religion,

    whereas

    the

    intrinsically motivated lives

    his

    religion"(p.434).

    As the descriptions in Table 1 indicate,

    these

    twotypesofreligiousness were originally

    considered to be ends of a bipolar continuum.

    But

    from the

    very

    beginningof theempirical

    research, doubtwascaston theappropriate-

    ness ofcharacterizing/ and

    E

    in thisway.

    Feagin (1964) reported a

    factor

    analysisin

    which items

    from

    / and

    E

    scales loadedon

    separate,

    orthogonal factors.Allport(1966b)

    himself

    began

    to

    take note

    of a

    group

    of

    "muddleheads

    who

    refuse

    to

    conform

    to our

    neat religious

    logic"

    (p. 6). These individuals

    agreed with items on both scales despite

    Allport's attempt to construct the scales to

    represent polar opposites.

    As a result ofthese findings,Allportex-

    panded hisoriginalbipolarapproach intoa

    fourfold typology. Those who agreed with

    items

    on the /scaleanddisagreedwithitems

    on the E scale he called intrinsics. Those

    who

    disagreed with

    /

    items

    and

    agreed

    with

    Eitemshecalledextrinsics.Thosewhoagreed

    with

    items

    on

    both scales

    he

    called

    indiscrim-

    inately proreligious

    (henceforth

    indiscrimi-

    nate)andthosewhodisagreed with itemson

    both

    scales

    he

    called

    indiscriminately

    antire-

    ligious

    (henceforth

    nonreligious).

    Although

    AllportandRoss(1967)originally developed

    a scoring procedurethatexcluded thenon-

    religiouscategory

    (on the

    presumption that

    therewould

    be a no nonreligious respondents

    in their

    sample

    of

    church members),

    Hood

    (1970) proposed classifying the

    four

    groups

    on thebasisofmedian splits. This procedure

    has since been

    followed

    by the majority of

    researchers

    using

    the

    typology.

    EarlyConceptual Critiques

    AsthedatesinTable 1indicate, thecon-

    ceptsof

    intrinsic

    andextrinsic religiousness

    underwenta

    rather long development before

    they

    were

    operationalized.

    When empirical

    studies addressingtheissues

    of

    / and began

    to bepublished, they were quickly followed

    by

    two articles that, on the basis of rather

    meager evidence, offered a preliminary as-

    sessment of the

    concepts.

    The first, by

    Hunt

    and

    King(1971),con-

    tendedthat

    E

    was auseful concept, butthat

    7

    was not

    because

    it was too

    "metaphysical"

    (p.354).Theotherreview,byDittes (1971),

    which appeared in the same issue of the

    Journal

    for the Scientific Study of

    Religion

    (JSSR), seems

    to have

    been based primarily

    onpurist, logical positivist viewsofscience

    and ad hominems against Allport. Dittes

    statedthat

    theintrinsic-extrinsic

    concept

    in-

    cluded a "heavy contraband load of value

    judgment that simply

    will

    not be

    sloughed

    off" (p.

    375). Dittes accused

    AUportof"ne-

    glectingthepurityofconceptualization" (p.

    380)

    and

    accused

    I-E

    researchers

    of

    "persis-

    tently neglect[ing] orresisting]" the devel-

    opment

    ofmultidimensional frameworks (p.

    381).

    The

    latter

    was

    manifestly

    not the

    case,

    becausein all three of theROSstudies up to

    that time, Allport

    and

    Ross (1967), Feagin

    (1964),andHood (1970)hadnoted that /

    andE

    were

    separate dimensions.

    Dittes (1971)also accused Allport

    of

    con-

    ceptualagglutination: that

    is,

    creating

    a new

    concept out of a constellation of independent

    concepts. This criticism

    cannotbe

    evaluated

    at a purely conceptual level, because whether

    a particular groupofconcepts does form a

    syndrome

    is an empirical issue. One should

    be free to posit such syndromes, solongas

    one is

    willing

    to

    proceed

    further and

    test

  • 7/25/2019 Donahue JPSP 1985 Religiosidad Metanlisis

    3/20

    402

    MICHAEL J. DONAHUE

    their

    validity.

    At the

    time

    of

    Dittes' review,

    there was not

    enough empirical evidence

    to

    addresstheissue.

    Since

    the

    publication

    of

    these

    early

    reviews,

    a

    considerable body

    ofl-E

    research

    has

    been

    produced. The purpose of this review is to

    apply

    those data

    to the

    followingquestions:

    1.

    Whatis therelation between/ and

    E?

    What factors influence the

    correlation

    be-

    tween

    them?

    3

    2.

    What are the correlates of the two

    constructs considered individually? What

    types of

    variables

    do

    each

    of the two

    con-

    structs

    form

    relations with,

    and what do these

    relations

    reveal concerning

    the

    nature

    of /

    and?

    3.

    What

    is theutilityof theAllport-Hood

    fourfold

    typology? Does

    it

    increase predictive

    power

    over

    the two

    unipolar constructs?

    4.

    Have later conceptual developments

    shedanyfurther lighton / andE?

    AnEmpiricalAssessment

    Method

    LiteratureSearch

    Thedata to be presented hereare based on areview

    oftheliterature publishedinEnglishto the end of 1982.

    It

    was conducted by means of (a) examining all volumes

    of

    JSSR,

    the

    Review of Religious Research,

    and the

    Journal of Psychology and Theology for

    relevant

    articles;

    (b) examining

    all the

    articles that

    the

    Social Science

    Citation Index listed as referencing Allport and Ross

    (1967)or

    Feagin

    (1964);and (c) the"ancestry"method:

    examining the reference lists of the articles

    obtained

    by

    the twoprevious

    methods.

    The general form of this review is that of a

    meta-

    analysis (Glass etal.,1981;Hunteretal., 1982),inwhich

    I

    seek

    to

    combine

    the

    results

    of

    independent studies

    in

    order to make empirical determinations of the issues

    generally

    addressed by literature reviews. In order to

    facilitatecombining

    the

    results

    of

    these studies, consid-

    eration

    was

    givenonly

    to

    studies

    inwhich / andE

    were

    defined

    intermsofeither Allportand

    Ross's(1967)

    ROS,

    or Feagin's (1964) Intrinsic-Extrinsic

    (l/E)

    Religiosity

    measure, which consistsalmostexclusively

    of a

    subset

    of

    theROS. Related

    approaches,

    suchasAllenand Spilka's

    (1967)

    "committed-consensual"

    dimensions and alter-

    native conceptualizations of intrinsic (Hoge, 1972) or

    extrinsic (Wilson, 1960)religiousness,were

    not

    included.

    Analyses

    The

    analyses

    of

    these

    data

    required

    a

    number

    of

    judgment calls

    (McGrath,

    Martin,

    Kulka,

    1982).The

    first

    involvedhow

    to

    calculatemean correlations across

    studies. Although

    Fisher's

    rto z transformation comes

    most

    quickly

    to mind, both Glass et al.

    (1981)

    and

    Hunter

    et al. (1982) contended that it may not be

    appropriate. The procedureadopted and

    reported

    here

    (consistent with the recommendations of Hunter et al.)

    was

    to

    weight

    the

    untransformed

    re bytheir

    sample size,

    sum

    them, and divide by the sum of the sample sizes.

    Weighted z

    scores

    were also computed, and though

    always

    greater,

    neverdiffered by

    more than

    .03from the

    means reported here.

    Asecond issueinvolved

    the

    choice

    of how to

    calculate

    or

    estimate theproduct-moment correlation between /

    and

    E

    when only contingency table

    data

    were

    reported.

    4

    After

    consultation with several sources (Carroll, 1961;

    Glass et al., 1981; Guilford, 1965; Nunnally, 1976;

    Taylor, 1972),

    both

    tetrachoric

    and phi

    coefficients

    were

    calculated.

    In

    light

    of

    comments

    by

    Guilford (1965,

    p.

    300), tetrachoric correlations were computed only on

    samples of 200 or more, and,whenused in computations,

    were

    weightedby 45% oftheir sample size. Examination

    of the marginal probabilities for the contingency

    tables

    indicated that the corrections for tetrachoric correlations

    recommended

    by

    Jenkins (1955;

    see

    also Fishman,

    1956)

    were

    not required. When phi coefficients were used in

    calculations, theywereweighted

    by

    their

    sample

    size.

    In

    no casedid themean correlations involvingphi coefficients

    differ

    bymore than.04

    from

    those involving tetrachorics.

    Results

    Relations

    Between

    I

    and

    E

    One

    way to

    address

    the

    relation

    between

    /

    andEis

    throughfactoranalysis

    of the

    items

    of

    the

    subscales.

    All five

    published reports

    of

    such

    factor analyses, Carey (1974), Elifson

    (1976), Feagin (1964), Patrick (1979), and

    Vincenzo,

    Hendrick,

    and

    Murray(1976),

    re-

    port similar

    findings:

    Items from

    the two

    subscales

    tend

    to

    load

    on two

    separate,

    or-

    thogonal

    factorswith few, ifany, cross-load-

    ings.

    (See also Batson & Ventis, 1982,

    5

    in

    which

    / andE

    tend

    to

    load

    on

    separate,

    orthogonal factors.)

    3

    Because of the bipolar emphasis inAllport'sconcep-

    tualization,

    / wasoriginallyreverse-scored,sothatagree-

    ment

    with

    the items resulted in a low

    scale

    score.For

    easeofconsideration, thesignsof allcorrelations reported

    here have been adjusted

    so

    that

    a

    higher

    score indicates

    agreement forboth scales.

    "In one case (Joe, McGee, & Dazey, 1977), exact

    reconstruction of the contingency

    table

    was

    based

    on

    reported degrees

    of

    freedom

    for

    Is

    inwhichthefour

    cells

    were

    compared.

    In two other

    cases

    (Alker & Gawin,

    1978; Dodrill, Bean, & Bostrom, 1973), information

    allowed reconstruction

    of

    tablesconsistentwith

    the

    data.

    'Batson

    and Ventis

    (1982)

    presented

    1-E

    data

    that

    was based on a pooling of several studies. Because this

    included otherwise unpublished data,

    the

    pooled

    data,

    rather than the individualstudies, have been usedhere.

  • 7/25/2019 Donahue JPSP 1985 Religiosidad Metanlisis

    4/20

    l-E

    META-ANALYSIS

    4

    Anotherway toaddressthe natureof the

    relationbetween / andEis to examinethe

    reportedcorrelations between them.InTable

    2 therelevant data arepresented, arranged

    from

    the

    most

    negative

    correlation

    to the

    mostpositive.The mean correlation across

    the 34 samples reported here is -.06. This

    mean,

    however,

    isstronglyinfluencedbyfour

    studies inwhicheitherthe Feagin I/E scale

    or factor-analytically refined versionsof the

    ROS

    were

    used; three of these studies had

    ratherlargesamplesizes.

    For

    ease

    of

    discus-

    Table

    2

    Intrinsic-Extrinsic Correlations for

    Various

    Samples

    Source

    Sample

    Dodrill

    etal.,

    1973

    Bolt,1977

    Strickland &

    Shaffer,

    1971

    Shoemaker&Bolt, 1977

    Spilka, 1977

    Batson,1976

    Paloutzian

    etal., 1978

    Kahoe,

    1974a

    Alker&Gawin, 1978

    Baker

    &

    Gorsuch, 1982

    Hood,

    1971

    Stewin&Anderson, 1974

    McConahay

    &

    Hough, 1973

    Kahoe

    & Dunn, 1975

    Allport&Ross,1967

    Minion& Spilka, 1976

    Thompson,

    1974

    Spilka, Stout, Minton,

    &

    Sizemore, 1977

    Spilka

    etal.,

    1968

    Batson & Ventis, 1982

    Paloutzianetal.,1978

    Thompson,

    1974

    Tate&Miller,

    1971

    Hood,

    1971

    Thompson, 1974

    Hoge&Carroll, 1973

    Elifson, 1976

    Hood,

    1978

    Hoge&Carroll, 1973

    Joeetal.,

    1977

    Elifson,

    1976

    Patrick, 1979

    Hood,1970

    Stewin, 1976

    Hunt

    &

    King,

    1971

    Mean r

    Mean

    r for ROS

    studies

    -.58

    -.57

    -.54

    -.45

    -.41

    -.41

    -.40

    -.37

    -.33"

    -.32

    -.31""

    -.29

    -.24

    -.22

    -.21

    -.19

    -.18"

    -.16

    -.16

    -.14

    -.13

    -.11

    -.11"

    -.08

    -.08

    -.04-

    -.01

    .00

    .01

    At

    .06

    b

    .09*

    .l l

    f

    .18

    .21

    ,24

    f

    -.06

    -.20

    evangelical

    Protestant

    students

    conservative

    Protestants

    college students

    conservativeProtestant students

    '"affiliated

    and

    active

    Christians"

    Princeton Seminary students

    adultSunday school and

    nonacademic

    college

    staff

    religiously

    conservative students

    "religiously active" church members, many

    denominations

    "religiouswilderness camping organization"

    Baptist psychology students

    eleventh-grade

    students

    Protestant graduate seminarians

    Catholics, Methodists,

    and

    Baptists

    manydenominations

    Protestant church members

    Catholic mothers

    respondents

    personally involved in religion

    college

    students

    "undergraduates with

    at

    least

    a

    moderate

    interest

    in

    religion"

    psychologystudents

    Catholic fathers

    Methodist laityandclergy

    Baptist students

    Catholic adolescents

    Methodist andPresbyterians, Southern

    sample

    Southern Baptist women

    psychology

    students

    Methodistsand

    Presbyterians,

    Northern

    sample

    psychology

    students

    Southern Baptist men

    Buddhists, Baptists, and Congregationalists

    psychlogy students

    eleventh-grade students

    four

    denominations

    N

    Total

    ROSN

    255

    62

    24

    51

    106

    67

    177

    333

    101

    52

    83

    107

    159

    70

    309

    67

    532

    167

    146

    258

    84

    532

    175

    83

    532

    343

    452

    147

    51 5

    167

    562

    91

    89

    100

    1356

    8271

    4952

    Note.

    Allstudies used AllportROSexceptas

    noted:Tetrachoric

    correlation.

    b

    Becauseofsample size, onlyphi

    coefficient

    calculated.

    c

    Not included in the calculation of the mean; same sample as.08correlation

    below.

    d

    Used

    Feagin

    //".

    e

    Correlations

    basedon

    stratified sample weighted

    to

    increase

    representativeness.

    f

    Scalesrefined

    through

    factoranalysis.

    I-E

    correlation reported

    inHunt&

    King,

    1971;

    sample

    and

    methodology

    reportedin

    King

    &

    Hunt,

    1972.

  • 7/25/2019 Donahue JPSP 1985 Religiosidad Metanlisis

    5/20

    4 4

    MICHAEL

    J. DONAHUE

    sion,only the 28 ROS correlations are con-

    sidered initially and the remaining studies

    later.

    Across

    the ROS

    studies,

    the

    mean corre-

    lation

    is

    .20,

    which

    is

    interesting because

    of itssimilarityto the .21originallyreported

    by

    Allport

    and

    Ross (1967).

    But

    more

    im-

    portant

    is the

    considerable range

    of the

    cor-

    relations across these studies: from .58 to

    .24. Is this variation due to theoperationof

    some moderator variable,

    or is it

    simply

    sampling

    variance? Hunter et al. (1982) pre-

    sented equations to address this question.

    Applying

    those equations to the ROScorre-

    lations

    in

    Table

    2, I

    found

    thatthecorrected

    (nonsampling)

    standard

    deviation

    for

    these

    datawas .15.

    The next question is, What moderator

    variable

    is involved? One possibility suggests

    itself on

    examination

    of the

    third column

    in

    Table2: thenatureof the samples. The four

    studiesdesribedby the researchers as involv-

    ing evangelical or conservative religionists

    have a

    mean

    of .44,

    which

    is significantly

    (p