Religiousness - Genetic And Enviromental Influences (1999)

Twin Research (1999) 2, 88–98 © 1999 Stockton Press All rights reserved 1369–0523/99 $12.00 http://www.stockton-press.co.uk/tr

I ntr i nsi c and extr i nsi c r el i gi ousness: geneti c and envi r onmental i nfl uences and per sonal i ty cor r el ates Thomas J Bouchard Jr, M att M cGue, Davi d Lykken and A uke Tel l egen University of Minnesota, Minneapolis, USA Thi s r epor t pr esents fi ndi ngs for the I ntr i nsi c (I R) and Extr i nsi c (ER) r el i gi ousness scal es fr om the M i nnesota Study of Tw i ns Rear ed A par t. The scal es w er e show n to be i nter nal l y consi stent, suffi ci entl y di sti nct fr om the scal es of the Cal i for ni a Psychol ogi cal I nventor y and the M ul ti di mensi onal Per sonal i ty Questi onnai r e and unr el ated to a number of measur es of r esponse styl e to justi fy tr eati ng them as di sti nct tr ai ts. The I scal es al so show ed consi der abl e evi dence of constr uct val i di ty i n i ts cor r el ati ons w i th r el i gi ous fundamental i sm and author i tar i ani sm as assessed by the M M PI and A l temeyer ’s Ri ght-Wi ng A uthor i tar i ani sm scal e. Data on I R and ER fr om 35 pai r s of monozygoti c tw i ns r ear ed apar t (M ZA ) and 37 pai r s of di zygoti c tw i ns r ear ed apar t (DZA ) w er e fi tted to a bi ometr i c model and demonstr ated si gni fi cant her i tabi l i ty (0.43 and 0.39), w i th a model contai ni ng geneti c pl us envi r onmental factor s fi tti ng si gni fi cantl y better than a model contai ni ng onl y an envi r onmental component. Tw i n si mi l ar i ty coul d not be expl ai ned by pl acement on a sel f-r epor ted measur e of fami l y M or al Rel i gi ous Emphasi s as measur ed by the Fami l y Envi r onment Scal e. Keyw or ds: tw i ns, rel i gi ousness, personal i ty

I ntr oducti on A ccordi ng to EO Wi l son, ‘The predi sposi ti on to rel i gi ous bel i ef i s the most compl ex and pow erful force i n the human mi nd and i n al l possi bi l i ty an i nerradi cal part of human nature.’ 1 Thi s evol uti onary theme has now been expl ored by a number of schol ars,2,3 al though i t appears to have been l argel y i gnored by evol uti onary psychol ogi sts.4,5 Wi l son’s cl ai m i s, of course, a rei terati on of si mi l ar asserti ons made over the mi l l enni a by thoughtful peopl e. Neverthel ess, i n spi te of the recogni ti on of i ts pervasi veness, compl exi ty and i mportance, rel i gi ousness as a psychol ogi cal trai t has been negl ected by modern psychol ogi sts rel ati ve to most other psychol ogi cal constructs. It has been rel egated, perhaps not to the fri nges, but to the si del i nes. Thi s i s surpri si ng gi ven the i mportance attached to rel i gi ousness by both Wi l l i am James,6 a foundi ng father of modern psychol ogy and the author of one of the di sci pl i ne’s most famous textbooks,7 and Gordon A l l port,8 the father of modern personal i ty theory. The most recent personal i ty textbooks to cross our desks9,10 do not contai n the terms rel i gi on or rel i gi ousness i n thei r i ndexes. Thomas and Carver 11 report si mi l ar fi ndi ngs based on thei r exami nati on of 60 textbooks devoted to chi l d and adol escent devel Correspondence: Thomas J Bouchard Jr, Department of Psychol ogy, El l i ott Hal l , Uni versi ty of M i nnesota, 75 East Ri ver Road, M i nneapol i s, M i nnesota 55455, USA . Fax: 612 6262079; E-mai l : [email protected] Recei ved 29 A pri l 1999; accepted 4 M ay 1999

opment. The fai l ure to ci te rel i gi ousness as an i mportant vari abl e i n major textbooks i s al so surpri si ng because rel i gi ousness i s a pow erful i nverse predi ctor of most of the debi l i tati ng condi ti ons psychol ogi sts purport to be i nterested i n understandi ng and al l evi ati ng – al cohol i sm, drug abuse, mental heal th, physi cal heal th, and del i nquency.12–14 These empi ri cal facts al one shoul d make rel i gi ousness a pri me candi date for extensi ve sci enti fi c anal ysi s. Our i nterest i n rel i gi ousness as a psychol ogi cal trai t, how ever, fl ow s from Wi l son’s cl ai m 1 that l i ght may be cast on the nature of thi s trai t i f w e pay attenti on to the soci obi ol ogy of rel i gi on: ‘By tradi ti onal methods of reducti on and anal ysi s sci ence can expl ai n rel i gi on but cannot di mi ni sh the i mportance of i ts substance’ (p 172). Unti l qui te recentl y most psychol ogi sts di d not even questi on the assumpti on that rel i gi ous atti tudes and bel i efs w ere l argel y shaped by parental i nfl uences. Soci ol ogi sts research the ‘i nheri tance’ of rel i gi ousness w i thout even menti oni ng the possi bi l i ty that genes may be i nvol ved.15 Even behavi oral geneti ci sts w ho have tested the assumpti on of common envi ronmental i nfl uence have been mi sl ed. Pl omi n,16 for exampl e, asserted: ‘Rel i gi osi ty, for exampl e, show s no geneti c i nfl uence and i s thus thought to be due to envi ronmental i nfl uence’ (p 97). Thi s cl ai m w as made on the basi s of data gathered on young peopl e w ho had not yet ‘l eft the nest’. It i s now know n that the effects of fami l y envi ronment di mi ni sh after chi l dren l eave thei r homes.17,18 Wi l son has recentl y made very cl ear w hat he means by tradi ti onal methods of reducti on

Intrinsic and extrinsic religiousness TJ Bouchard et al 89

and anal ysi s. Speci fi cal l y they i ncl ude repl i cati on, mensurati on, parsi mony, heuri sti cs and consi l i ence.19 We bel i eve a few comments on the fi rst four as they appl y to rel i gi ousness w oul d be w orthw hi l e. The l ast, consi l i ence, i s unquesti onabl y rel evant but beyond the scope of thi s paper. Replication It i s now reasonabl y w el l establ i shed that vari ati on i n rel i gi ousness i s, contrary to w i despread bel i ef, moderatel y i nfl uenced by geneti c factors.20 Thi s fi ndi ng throw s al l studi es of the i nfl uence of envi ronmental vari abl es, based on studi es of bi ol ogi cal fami l i es, i nto seri ous doubt, as such desi gns assume rather than test for envi ronmental i nfl uences. The fi ndi ng of geneti c i nfl uence on any trai t rai ses an enti re array of new questi ons, for exampl e, w hat i s, a) the degree of addi ti ve and/ or nonaddi ti ve geneti c i nfl uence, b) the nature of the envi ronmental i nfl uences (shared or unshared), c) the rol e of genes and envi ronment on devel opmental change, d) the i nfl uence of assortati ve mati ng, etc. We al ready have some i nkl i ngs regardi ng these i nfl uences, but i t shoul d be cl ear that such fi ndi ngs must al so be subjected to constructi ve repl i cati on. Constructi ve repl i cati on i nvol ves the use of al ternate desi gns and tool s of measurement to demonstrate that a fi ndi ng i s i ndependent of the uni que features of a parti cul ar desi gn and measurement tool .21 Thi s study addresses both of these i ssues as w e uti l i ze tw i ns reared apart, a rare desi gn, and empl oy the Intri nsi c (IR) and Extri nsi c (ER) measures of rel i gi ousness, scal es not previ ousl y used i n behavi or geneti c studi es. A ddi ti onal desi gns, not i nvol vi ng tw i ns, especi al l y studi es of unrel ated i ndi vi dual s reared together as chi l dren and fol l ow ed up as adul ts, w oul d be very desi rabl e. That geneti c factors i nfl uence rel i gi ousness i s a suffi ci entl y counter-i ntui ti ve fi ndi ng that i t w i l l be accepted as an empi ri cal fact onl y after i t has been constructi vel y repl i cated numerous ti mes. Mensuration Constructi ve repl i cati on requi res the exi stence of adequate measuri ng i nstruments. The probl em of adequate measurement pl agues the study of rel i gi ousness. M ore and better measuri ng i nstruments are i ndi spensabl e for advance to occur. When measures of rel i gi on/ rel i gi ousness are i ncl uded i n studi es, they often tend to be si mpl e nonfocal vari abl es such as denomi nati onal membershi p, frequency of attendance at rel i gi ous servi ces, or a rati ng of the i mportance of rel i gi on to the i ndi vi dual . M ore speci fi c scal es often focus onl y on Chri sti an orthodoxy rather than attempt to capture a common

component of the di verse bel i ef systems i n most of the popul ati ons actual l y bei ng studi ed.22 Better and more di verse i nstrumentati on i s sorel y needed. The study reported i n thi s paper exami nes geneti c and envi ronmental i nfl uences on Intri nsi c (IR) and Extri nsi c (ER) rel i gi ousness. The di sti ncti on betw een these tw o measures i s a fundamental one i n the domai n of the psychol ogy of rel i gi on.13,23 The di sti ncti on w as best summed up by A l l port and Ross24 w ho stated that ‘the extri nsi cal l y moti vated person uses hi s rel i gi on, w hereas the i ntri nsi cal l y moti vated person l i ves hi s rel i gi on’ (p 434). The IR and ER scal es di sti ngui sh betw een i nstrumental val ues (Extri nsi c Rel i gi ousness) as a mode of conduct and termi nal val ues (Intri nsi c Rel i gi ousness) as an end state of exi stence. Thi s di sti ncti on al so underl i es Rokeach’s w i del y used i nstrument, the Val ue Survey,25 i n w hi ch i nstrumental and termi nal val ues consti tute separate l i sts that are ranked i ndependentl y of each other. When w e deci ded to i ncl ude a measure of rel i gi ousness i n the M i nnesota Study of Tw i ns Reared A part (M ISTRA ), a number of col l eagues suggested to us that many measures of rel i gi ousness l argel y refl ect response sets – soci al l y desi rabl e responses and acqui escence – and/ or perhaps a mani pul ati ve/ soci al use of rel i gi on, not ‘true’ or ‘i ntri nsi c’ rel i gi ousness. A search of the l i terature l ed us to the A ge Uni versal Rel i gi ous Ori entati on Scal e (A UROS).26 Thi s i nstrument i s a modi fi ed versi on of the A l l port and Ross I–E scal es.24 Gorsuch and Venabl e revi sed the I–E scal e to al l ow i ts use w i th chi l dren and young adol escents by rew ri ti ng i tems to match a fi fth grade readi ng l evel . Thi s l atter feature w as i mportant to us because of the extremel y w i de range of educati onal background and readi ng abi l i ty among the parti ci pants i n M ISTRA . For exampl e, numerous parti ci pants have found the Study of Val ues27 very di ffi cul t and confusi ng because of the compl exi ty l evel of the questi ons and the col l ege l evel content. There i s al so good evi dence that soci al desi rabi l i ty i s not a probl em w i th vari ous measures of IR and ER.13,23,28 Our tw i ns had al ready compl eted the Cal i forni a Psychol ogi cal Inventory (CPI) and the M ul ti di mensi onal Personal i ty Questi onnai re (M PQ). Both i nstruments contai ns a number of val i di ty scal es. Thi s al l ow ed us to eval uate di rectl y the possi bl e rol e of soci al desi rabi l i ty and acqui escence i n thi s sampl e, as w el l as exami ne the IR and ER scal e correl ati ons w i th w el l establ i shed personal i ty i nventori es. On the CPI the Good Impressi on (Gi ) scal e measures the tendency to pai nt an excessi vel y posi ti ve pi cture of one’s sel f. The Di cken Soci al Desi rabi l i ty scal e (DSD) i s al so a measure of soci al desi rabi l i ty and the Di cken A cqui escence scal e (DA C) i s a measure of yea sayi ng. Strong posi ti ve correl ati ons betw een these scal es and the IR and ER

Intrinsic and extrinsic religiousness TJ Bouchard et al 90

scal es w oul d throw doubt on the useful ness of the A UROS as a measure of rel i gi ousness. The M PQ contai ns a number of val i di ty scal es one of w hi ch, Unl i kel y Vi rtues, al so assesses the soci al desi rabi l i ty response set. Parsimony One reason psychol ogi sts have not studi ed rel i gi ousness i n more detai l i s because many of them bel i eve that i t i s al ready subsumed by major personal i ty trai ts. The evi dence si mpl y does not support thi s concl usi on. The correl ati ons betw een measures of rel i gi ousness and personal i ty are general l y smal l 20 and, w hen they exi st, di fferent personal i ty measures correl ate di fferenti al l y w i th di fferent measures of rel i gi ousness. A l temeyer,29 for exampl e, has show n that IR and ER correl ate di fferenti al l y w i th the Ri ght Wi ng A uthori tari ani sm scal e (RWA ). IR correl ates 0.36 and 0.41 w i th RWA i n student (n = 406) and parent (n = 549) sampl es. The correl ati ons w i th ER w ere –0.10 and –0.09. He presents i ntercorrel ati ons betw een a w i de range of rel i gi ousness measures (as w el l as some scal es rel ated to rel i gi ous atti tudes and bel i efs created for hi s studi es) and RWA . A l l of the val i di ty coeffi ci ents for IR are i n the 0.55 to 0.65 range, w hereas the correl ati ons for ER are negati ve and i n the –0.15 to –0.30, range confi rmi ng the concl usi on that IR does i ndeed have construct val i di ty and that ER measures a di fferent ki nd of rel i gi ousness at best. The correl ati ons found i n the parent and student sampl es are essenti al l y the same, thus provi di ng a repl i cati on across generati ons. There i s some suggesti on that correl ati ons betw een rel i gi ousness measures and personal i ty di ffer somew hat from one rel i gi ous group to another.30 Such fi ndi ngs, how ever, may si mpl y refl ect the use of smal l sampl es and requi re repl i cati on w i th much l arger sampl es. There i s no support for the argument that rel i gi ousness shoul d be excl uded from the armamentari um of i ndi vi dual di fference measures on the grounds of parsi mony. On the contrary, the l ack of ci tati on of rel i gi ousness i n most textbooks show s that the questi on of parsi mony has hardl y begun to be addressed. The l ocati on of vari ous rel i gi ousness measures i n the domai n of psychol ogi cal vari abl es necessary to cover comprehensi vel y the enti re array of meani ngful human i ndi vi dual di fferences has yet to be determi ned and deserves a great deal more attenti on. Heuristics The i ssue of heuri sti cs has al ready been touched upon. Rel i gi ousness i s an i mportant construct because i t i s si gni fi cantl y and substanti vel y rel ated

to many real -l i fe condi ti ons of i ntense i nterest to soci al sci enti sts. Those w ho see rel i gi ousness as a posi ti ve trai t emphasi ze i ts predi cti ve si gni fi cance for mental and physi cal heal th behavi or. On the other hand, those w ho see i t as a mal evol ent force rel ate i t to such vari abl es as Ri ght Wi ng A uthori tari ani sm, Dogmati sm and other ‘negati ve atti tudes’ and l argel y i gnore i ts posi ti ve aspects.29,31 Rel i gi ousness appears to be a psychol ogi cal l y compl ex trai t that l i es at the center of a nexus of correl ati ons i nvol vi ng, among other thi ngs, i mportant l i fe styl e choi ces i nfl uenci ng physi cal heal th, mental states i nfl uenci ng mental heal th and i mportant atti tudi nal vari abl es (some posi ti ve, some negati ve). A thorough understandi ng of the vari ous facets of rel i gi ousness and thei r causal antecedents w oul d move us a l ong w ay tow ard understandi ng the genesi s of many real l i fe probl ems. Thi s study al so al l ow s a test of an addi ti onal hypothesi s about IR and ER. In a previ ous study of geneti c i nfl uence on measures of rel i gi ousness (not the IR and ER scal es used i n thi s study), usi ng M ZA and DZA tw i ns, w e found that some of the DZA tw i n correl ati ons w ere near zero suggesti ng the trai t mi ght i nvol ve a confi gural , nonaddi ti ve combi nati on of geneti c effects.32 We predi ct si mi l ar fi ndi ngs here.

M ethod Subjects The reared-apart tw i ns w ere parti ci pants i n M ISTRA betw een the years of 1979 and 1998. Detai l s of thei r recrui tment and testi ng are found el sew here.33,34 To ensure comparabl e sampl es of adul t parti ci pants across publ i cati ons, no parti ci pants under 18 years of age are i ncl uded i n the sampl e used i n thi s paper. Tabl e 1 presents the descri pti ve characteri sti cs of the sampl e. Note that the A UROS and RWA scal es w ere mai l ed out to many of the tw i ns (see bel ow ), so the fi gures gi ven i n Tabl e 1 descri bed the sampl e at the ti me of i ntake assessment onl y. Femal es are over-represented as they are i n most tw i n studi es.35 The tw i ns are pri mari l y mi ddl e-aged adul ts, and have experi enced varyi ng degrees of separati on and contact. Zygosi ty for al l pai rs w as determi ned by the M i nneapol i s M emori al Bl ood Bank (M M BB) w hi ch has changed the geneti c systems assessed for determi nati on of tw i n type i n recent years. Unti l 1994, M M BB never used few er than ni ne protei n markers (bl ood groups, serum protei ns and enzymes). M M BB currentl y uses si x bl ood group markers and three DNA pol ymorphi sms. The probabi l i ty that a DZA pai r w oul d be concordant on al l markers and thus mi scl assi fi ed as an M ZA pai r i s l ess than 0.001. One tw i n pai r w as di agnosed on the basi s of observati on by study

Intrinsic and extrinsic religiousness TJ Bouchard et al 91 Tabl e 1 M eans and standard devi ati ons for age at assessment, measures of Separati on and Contact and percentage of femal es for M ZA and DZA tw i n sampl es

M ZA (n = 35) M ean SD DZA (n = 37) M ean SD

Time Together Prior to Separation (days)

Total Contact Time (weeks)

Time Apart (months)

65.7

151.1 244.2

68.4 116.7

415.7 202.4

56.8

287.1 390.9

46.2 55.1

528.3 171.4

Age (years)

% Females

41.9 12.2 47.2 14.2

i nvesti gators because thei r bl ood sampl e w as unusabl e upon reachi ng the M M BB. Whenever possi bl e, spouses and partners and i n some cases chi l dren and fri ends of tw i ns parti ci pate i n the study thus provi di ng i nformati on on assortati ve mati ng and a l arge sampl e for age and sex correcti on of the data (see bel ow ) and exami nati on of the correl ati ons betw een i nstruments based on a sampl e of adequate si ze. Instruments A mong the many i nstruments compl eted by the parti ci pants i n M ISTRA the fol l ow i ng w ere used i n thi s paper. The 480-i tem versi on of the Cal i forni a Psychol ogi cal Inventory (CPI),36 the M i nnesota M ul ti phasi c Personal i ty Inventory (M M PI),37 the M ul ti di mensi onal Personal i ty Questi onnai re (A Tel l egen, 1982, unpubl i shed manuscri pt), the Fami l y Envi ronment Scal e,38 a modi fi ed versi on of the A UROS and the Ri ght-Wi ng A uthori tari an scal e (RWA ).29 Because of factors such as age, i nfi rmi ti es, ti me avai l abl e for assessment, not al l tw i ns w ere abl e to compl ete al l components of the assessment. The A UROS and the RWA w ere i ncorporated i nto the assessment battery l ate i n the study and at di fferent ti mes. A t the ti me of i ncorporati on they w ere mai l ed to the previ ous tw i n parti ci pants, w ho w e w ere confi dent w oul d compl ete them i n a rel i abl e fashi on. Both scal es have been admi ni stered to subsequent tw i ns as part of the regul ar assessment. We modi fi ed the A UROS somew hat to accommodate rel i gi ons other than those from a Judeo-Chri sti an heri tage. The standard response format i s a fi ve poi nt Li kert scal e runni ng from strongl y di sagree to strongl y agree. We modi fi ed thi s format to be consi stent w i th other i nstruments i n our battery (–4 = Very strongl y di sagree, –3 = Strongl y di sagree, –2 = M oderatel y di sagree, –1 = Sl i ghtl y di sagree; 0 = Exactl y and preci sel y neutral etc to + 4 = Very strongl y agree). M ai ntai ni ng a consi stent response format over a number of di fferent i nstruments reduces confusi on among the many parti ci pants i n our study w ho have never before compl eted psychol ogi cal i nventori es. Consi stent w i th previ ous procedures w e provi ded the parti ci pants w i th the opti on of respondi ng ‘not

appl i cabl e’ (n/ a). Respondents used the n/ a response rather l i beral l y and i t appeared that w e w oul d l ose many cases due to there bei ng too few i tems on the al ready bri ef scal es. Upon questi oni ng parti ci pants i n M ISTRA , w e found that they al most al w ays used n/ a to mean not sure one w ay or the other. Thi s i s the meani ng of the mi ddl e or neutral response accordi ng to Gorsuch and Venabl e.26 Non-responses are a common probl em w hen usi ng the I–E scal es w i th sampl es contai ni ng a si zeabl e proporti on of nonrel i gi ous parti ci pants. The standard procedure i s to count non responses as neutral responses.23,29 We fol l ow ed that procedure here, scori ng n/ a responses as 0. Intri nsi c (IR) or Extri nsi c (ER) scores w ere deri ved fol l ow i ng Gorsuch and Venabl e,26 taki ng i nto account the i tem keyi ng correcti on reported by Gorsuch.13 The revi sed i tems used i n thi s study are show n i n Tabl e 2 Items 1, 5, 6, 8, 10, 11, 15, 18 and 20 consti tute the IR scal e, and i tems 2, 3, 4, 7, 9, 12, 13, 14, 16, 17, and 19 consti tute the ER scal e.

A nal yti c pr ocedur es Age and sex Scores on each of the scal es w ere corrected for age and sex usi ng procedures descri bed by M cGue and Bouchard 39 based on al l the scores avai l abl e (n = 253). Analysis of the infl uence of separation and contact Reared-apart tw i n studi es are not trul y randomi zed experi ments because the tw i ns vary i n thei r degree of separati on and thei r degree of contact. A s i ndi cated i n Tabl e 1, these tw i ns w ere separated at vari ous ages and had varyi ng degrees of contact pri or to parti ci pati on i n the M ISTRA assessment. The si mi l ari ty of tw i ns reared apart on vari ous psychol ogi cal trai ts has someti mes thought to be due to contact betw een the tw i ns.40,41 In order to test thi s hypothesi s w e gathered rel evant data. The separati on/ contact measures i n Tabl e 1 w ere based on detai l ed i ntervi ew s w i th both tw i ns and a revi ew of

Intrinsic and extrinsic religiousness TJ Bouchard et al 92 Tabl e 2 Vari max factor l oadi ngs of the Revi sed A UROS i tems for a tw o factor sol uti on (n = 253). Loadi ngs w hose absol ute val ue w as l ess than 0.30 are omi tted Item number, text of item and scale on which the item is scored (IR or ER) 10. 1. 5. 20. 6. 8. 15. 18. 3. 11. 9. 12. 19. 14. 16. 2. 7. 17. 4. 13.

M y rel i gi on i s i mportant because i t answ ers many questi ons about the meani ng of l i fe. (IR) I enjoy readi ng about my rel i gi on. (IR) It i s i mportant to me to spend ti me i n pri vate thought and prayer. (IR) a I w oul d prefer to go to my pl ace of w orshi p . (IR) I have often had a strong sense of God’s presence. (IR) I try hard to l i ve al l my l i fe accordi ng to my rel i gi ous bel i efs. (IR) M y w hol e approach to l i fe i s based on my rel i gi on. (IR) Prayers I say w hen I’m al one are as i mportant to me as those I say i n my pl ace of w orshi p. (IR) It doesn’t much matter w hat I bel i eve as l ong as I am good. (ER) I w oul d rather joi n a rel i gi ous study group (a group that studi es the Bi bl e, Koran, Torah or other rel i gi ous text) than a soci al group at my pl ace of w orshi p (Church, Synagogue, Templ e). (IR) What rel i gi on offers me most i s comfort i n ti mes of troubl e and sorrow. (ER) Prayer i s for peace and happi ness. (ER) A l though I bel i eve i n my rel i gi on, many other thi ngs are more i mportant i n l i fe. (ER) I go to my (Church, Synagogue, Templ e) to spend ti me w i th my fri ends. (ER) I go to my (Church, Synagogue, Templ e) mai nl y because I enjoy seei ng peopl e that I know there. (ER) I go to my pl ace of w orshi p (Church, Synagogue, Templ e) because i t hel ps me to make fri ends. (ER) I pray mai nl y to gai n rel i ef and protecti on. (ER) I pray mai nl y because I have been taught to pray. (ER) Someti mes I have to i gnore my rel i gi ous bel i efs because of w hat peopl e mi ght thi nk of me. (ER) A l though I am rel i gi ous, I don’t l et i t affect my dai l y l i fe. (ER)

Percent Vari ance A ccounted for

IR

ER

0.77 0.75 0.74 0.74 0.73 0.72 0.71 0.68 –0.63 0.55 0.53 0.44 –0.39

0.36 0.80 0.79 0.66 0.56 0.50 0.42

30.0

14.1

a

Response format for i tem 20. a) a few ti mes a year or l ess, b) once every month or tw o, c) tw o or three ti mes a month, d) about once a w eek, e) more often than once a w eek, f) about once a day.

avai l abl e records. The i nfl uence of these vari abl es w as eval uated by correl ati ng each measure w i th tw i n pai r absol ute di fferences on the IR and ER scal es. If tw i n contact accounts, at l east i n part, for tw i n si mi l ari ty, then tw i n pai r di fferences shoul d be negati vel y correl ated w i th i ndi ces of ti me spent together and posi ti vel y correl ated w i th ti me spent apart. Because the A UROS and the RWA w ere mai l ed to many of the tw i ns after thei r parti ci pati on i n the assessment (l ong after i n some cases and shortl y thereafter i n others) the measures of Total Contact and Ti me A part are onl y rough approxi mati ons.

bi ol ogi cal correl ati on confounds geneti c and envi ronmental i nfl uences, w hereas the adopted correl ati on i s uncontami nated by geneti c i nfl uence. To esti mate the i nfl uence of M RE on the M ZA i ntracl ass correl ati on for IR, w e need onl y mul ti pl y the magni tude of pl acement by the square of the rel evant envi ronmental correl ati ons (0.32
(0.10
0.10)). Under the hypothesi s of no geneti c i nfl uence the predi cted M ZA correl ati on for I w oul d be 0.003. The correl ati on betw een M RE and E i n both the adopted and bi ol ogi cal groups i s sl i ghtl y negati ve and near zero. Pl acement, at l east on M RE, does not appear to expl ai n much tw i n si mi l ari ty on IR and ER.

Analysis of the infl uence of self-reported family rearing environment

Model fi tting

The M oral Rel i gi ous Emphasi s (M RE) scal e of the FES i s the theoreti cal l y most rel evant possi bl e predi ctor of I and E i n thi s data set. If i t i s a trai trel evant causal vari abl e and i f the tw i ns have undergone pl acement on thi s vari abl e, then i ts i nfl uence may be i mportant. The i ntracl ass correl ati on for the age and sex-corrected scores on the M RE scal e for the M ZA tw i ns (n = 68 pai rs) i s 0.32, suggesti ng a moderate amount of pl acement. The correl ati on betw een M RE and I for M ISTRA parti ci pants reared by unrel ated i ndi vi dual s i s, how ever, onl y 0.10 (n = 127) suggesti ng at best a very w eak causal i nfl uence. The correl ati on for parti ci pants reared by bi ol ogi cal fami l i es i s 0.53 (n = 42). The

The quanti tati ve geneti c model assumes that observed phenotypi c vari ance (V p ) i s a l i near addi ti ve functi on of geneti c (V g) and envi ronmental (Ve) vari ances. Symbol i cal l y, V p = V g + Ve. A s our tw i ns have been reared apart, the envi ronmental vari ance i n the model represents resi dual vari ance not expl ai ned by geneti c i nfl uence, i e nonshared envi ronmental vari ance confounded w i th measurement error. From quanti tati ve geneti c theory and under the assumpti on that al l geneti c effects are addi ti ve (Va = vari ance due to addi ti ve gene effects), w e can deri ve the expected covari ance betw een any tw o

Intrinsic and extrinsic religiousness TJ Bouchard et al 93

rel ati ves as a functi on of the vari ance components gi ven above. The expected covari ances betw een the M ZA and DZA tw i n pai rs w i l l be COV COV

(M ZA ) (DZA )

= Va = 0.5Va

The general assumpti ons i n the model are: (a) there i s no genotype–envi ronment correl ati on or i nteracti on, (b) al l geneti c effects are addi ti ve effects, (c) mati ng i s random w i th regard to the trai ts under study, and (d) no sel ecti ve pl acement on trai t rel evant factors has occurred. We test the assumpti on of random mati ng bel ow. Extensi ve di scussi ons of the other assumpti ons i n behavi oral geneti c desi gns are avai l abl e el sew here.42 Expected vari ances and covari ances based on the model for the M ZA and DZA tw i ns w ere appl i ed to the observed vari ances and covari ances usi ng a maxi mum l i kel i hood esti mati on procedure i n M x.43 One of the advantages of the model -fi tti ng techni que i s that i t provi des the opportuni ty to test competi ng theoreti cal model s. We fi tted a ful l model and a reduced model to the tw i n covari ances. The ful l model i ncl uded the addi ti ve geneti c vari ance component and the unshared envi ronment/ measurement error vari ance component; the reduced model entai l ed droppi ng the addi ti ve geneti c vari ance component. To measure overal l fi t of the model , w e used the χ2 test stati sti c w i th the cri teri on for rejecti on of model s at the 0.05 probabi l i ty l evel . In eval uati ng the rel ati ve fi ts of the vari ous model s, parti cul arl y those that coul d not be rejected on the basi s of the χ2 test stati sti c al one, the A kai ke Informati on Cri teri on [A IC = χ2 – 2(df)] w as exami ned. A IC quanti fi es the i nformati on content of a model i n terms of the joi nt cri teri a of fi t and parsi mony. In general , smal l χ2 val ues from model s w i th few free parameters l ead to smal l A ICs, representi ng maxi mum parsi mony, w hereas l arge χ2 val ues from model s w i th many parameters yi el d l arge A ICs, representi ng l ack of parsi mony.44

Resul ts The al pha rel i abi l i ti es, based on 253 cases, w ere 0.88 for IR and 0.66 for ER. The IR i tems cl earl y form a coherent and i nternal l y consi stent scal e. The ER scal e i s much l ess i nternal l y consi stent. The correl ati on betw een IR and ER for al l cases (n = 253) w as 0.04, confi rmi ng that i n a rel i gi ousl y heterogeneous sampl e the measures are uncorrel ated. A factor anal ysi s of the i tems yi el ded three ei genval ues greater than 1.00 and a readi l y i nterpretabl e tw o factor sol uti on, al though there w as a smal l number of i tems that l oaded on both factors. The resul ts of the tw o factor sol uti on are gi ven i n

Tabl e 2. The factor scores correl ated 0.97 (IR) and 0.93 (ER) w i th the scal es. We chose to anal yze the standard scal es rather than factor scores i n order to al l ow a more di rect compari son of our fi ndi ngs w i th those of other studi es that have used the IR and ER scal es. The means and standard devi ati ons for the IR and ER scal es, by zygosi ty, are show n i n Tabl e 3. The M ZA and DZA tw i ns do not di ffer si gni fi cantl y w i th respect to means or vari ances. The correl ati ons betw een absol ute di fferences betw een members of each tw i n pai r on IR and ER and separati on and contact are gi ven i n Tabl e 4 for both types of tw i n. None of the correl ati ons i s si gni fi cant for the M ZA tw i ns for ei ther measure. A l l three measures, how ever, correl ated si gni fi cantl y w i th ER i n the DZA sampl e and the correl ati ons w i th IR, w hi l e not stati sti cal l y si gni fi cant, are si mi l ar. The correl ati ons are al l i n the predi cted di recti on. Ti me rai sed apart correl ates posi ti vel y w i th absol ute di fference and amount of contact correl ates negati vel y. Total Contact Ti me (TCT) i ncorporates Ti me Together Pri or to Separati on (TTPS). Contact after Separati on (TCT–TTPS) has no correl ati on w i th ei ther IR (0.07) or ER (–0.04); consequentl y al l the correl ati on i s accounted for by the measure of earl y contact. Ti me A part i s correl ated –0.10 w i th TTPS; consequentl y i t contri butes i ndependentl y to the predi cti on of the di fference score. The adjusted mul ti pl e r for predi cti ng ER i s 0.42. The i ntracl ass correl ati ons and 95% confi dence i nterval s for the IR and ER scal es for the M ZA and DZA tw i ns are gi ven i n Tabl e 5. The correl ati ons for IR show a pattern that i s typi cal for many personal i ty vari abl es, w i th the DZA correl ati on (0.20) bei ng about hal f the M ZA correl ati on (0.37). ER show s a much l ess regul ar pattern, w i th the DZA correl ati on (0.38) bei ng l arge than the M ZA correl ati on (0.24). The i ntercl ass correl ati ons for spouses based on 40 pai rs (i ncl udi ng some fol l ow -up pai rs not used i n thi s study) for the age and sex corrected scores w ere 0.32 for IR and 0.13 for ER. These assortati ve mati ng coeffi ci ents are suffi ci entl y l ow that, gi ven the modest sampl es of M ZA and DZA tw i ns, w e chose not to i ncl ude assortati ve mati ng i n our model i ng. The resul ts of model fi tti ng the vari ance–covari ance matri ces for the age and sex corrected IR and ER scores are show n i n Tabl e 6. Consi stent w i th w hat one w oul d i nfer from the i ntracl ass correl ati ons, the Tabl e 3 M eans and standard devi ati ons for the Intri nsi c and Extri nsi c M easures of Rel i gi ousness for M ZA and DZA tw i ns Scale Intrinsic Religiousness (IR) Extrinsic Religiousness (ER)

MZA (n = 70) Mean SD

DZA n = 74) Mean SD

47.7

15.7

47.8

14.3

51.6

11.3

54.1

13.5

Intrinsic and extrinsic religiousness TJ Bouchard et al 94 Tabl e 4 Correl ati ons betw een the absol ute di fferences i n Intri nsi c (IR) and Extri nsi c (ER) Rel i gi ousness scores and measures of Contact and Separati on for M ZA and DZA tw i ns Contact measures

M ZA (n = 35) DZA (n = 37) IR ER IR ER

Ti me Together Pri or to Separati on (days) Total Contact Ti me (w eeks) Ti me A part (months)

0.09 –0.15 –0.05

–0.12 –0.11 –0.04

–0.27 –0.27 0.19

–0.33a –0.33a 0.37a

aP<0.05

general model fi tted the data much better for both IR and ER than a purel y envi ronmental model , even though the l atter coul d not qui te be rejected for ei ther vari abl e. The χ2 for the general model are smal l as are the A IC i ndi ces i ndi cati ng a good fi t. When the geneti c parameter i s dropped from the model there i s a si gni fi cant i ncrease i n the χ2 stati sti c for both the IR (χ2 = 8.23, (1(df), P < 0.01) and ER (χ2 = 6.06, 1(df), P < 0.01) scal es i ndi cati ng poorer fi t. The standardi zed vari ance esti mates and 95% confi dence i nterval for the tw o parameters for both vari abl es are gi ven i n Tabl e 5. The correl ati on betw een the IR and ER scal es and the CPI scal es are show n i n Tabl e 7. A l though a few of the correl ati ons reach stati sti cal si gni fi cance due to the l arge sampl e si ze, they are al l very modest. Because of the l arge number of correl ati ons w e onl y di scuss those si gni fi cant at P > 0.01. Sel f-Control , Femi ni ni ty/ M ascul i ni ty, Responsi bi l i ty and Good Impressi on correl ate 0.21, 0.21, 0.20 and 0.18 w i th IR. Sel f-A cceptance, Independence and Domi nance have negati ve correl ati ons of –0.21, –0.20 and –0.19 w i th ER. There are no si gni fi cant correl ati ons betw een ei ther IR or ER and the Vector scal es w hi ch refl ect the core theoreti cal constructs that purportedl y underl i e covari ati on i n the scal es of the CPI. Of parti cul ar i nterest i s the vi rtual l ack of any si gni fi cant correl ati on betw een the Di ckens measures of response styl e (DSD and DA C) and IR and ER. A s noted above, Good Impressi on, how ever, has a correl ati on of 0.18 w i th IR. The correl ati ons betw een the M PQ scal es and the IR and ER scal es are gi ven i n Tabl e 8. There i s a negati ve correl ati on betw een A ggressi on and IR (–0.30) and a posi ti ve correl ati on betw een Tradi ti onal i sm and IR (0.37). The hi gher order Constrai nt factor scal e correl ates 0.32 w i th IR. The l argest M PQ

correl ate of ER i s 0.15 for the Harm Avoi dance scal e. Nei ther IR nor ER correl ates si gni fi cantl y w i th any of the response styl e i ndi ces of the M PQ, i ncl udi ng the Unl i kel y Vi rtues scal e w hi ch i s a measure of soci al desi rabl e respondi ng, and True Response Inconsi stency w hi ch i s a content free measure of acqui escence. In order to compare our sampl e w i th A l temeyer’s, w e correl ated hi s measure of RWA (w hi ch correl ates 0.71 w i th the M PQ Tradi ti onal i sm scal e) w i th IR and ER. The correl ati ons (n = 171) w ere 0.47 and –0.10 for IR and ER, respecti vel y. Si nce IR correl ates 0.47 w i th RWA and 0.37 w i th Tradi ti onal i sm i t i s cl ear that facets of IR overl ap w i th the construct of Tradi ti onal i sm. We al so computed the correl ati on betw een the M M PI Rel i gi ous Fundamental i sm (REL) scal e used i n the Wal ter et al 32 study w i th the IR and ER scal es. The correl ati ons (n = 195) w ere 0.68 and –0.22.

Di scussi on The al pha rel i abi l i ti es for IR (0.88) and ER (0.66) are comparabl e w i th those found i n other adul t sampl es. In A l temeyer’s29 student sampl e the val ues w ere 0.86 and 0.56, and i n hi s parent sampl e they w ere 0.89 and 0.72. Gorsuch 26 reports al phas of 0.73 and 0.66 for an adul t rel i gi ous (members of a church) sampl e (n = 101). The ER scal e i s cl earl y l ess i nternal l y consi stent than i s desi rabl e and coul d probabl y be i mproved w i th addi ti onal i tem refi nement. The correl ati on of 0.04 betw een the IR and ER scal es show s cl ear di scri mi nant val i di ty and i s reasonabl y si mi l ar to the val ue of –0.16 reported by Donahue23 i n hi s meta-anal ysi s of IR and ER correl ati ons for non-rel i gi ous sampl es. Note that si nce al l the IR and ER i tems are stated i n a posi ti ve fashi on, acqui escent response set shoul d resul t i n a posi ti ve correl ati on betw een the tw o scal es. These resul ts confi rm the vi ew that IR and ER assess di fferent di mensi ons of rel i gi ousness i n unsel ected popul ati ons and justi fi es the use of tw o measures of rel i gi ousness rather than one. The sl i ght rew ordi ng of the i tems and the expanded response format used i n thi s study does not appear to have changed the psychometri c properti es of the scal es.

Tabl e 5 Intracl ass correl ati ons for M ZA and DZA tw i ns and standardi zed vari ance esti mates for the A E model w i th 95% confi dence i nterval s for the Intri nsi c and Extri nsi c Rel i gi ousness Scal es Scale Intrinsic Religiousness (IR) Extrinsic Religiousness (ER)

Intraclass correlations MZA DZA 0.37 (0.04 – 0.63) 0.24 (–0.10 – 0.53)

0.20 (–0.13 – 0.49) 0.38 (0.06 – 0.63)

Variance estimates VA

VE

0.43 (0.15 – 0.64) 0.39 (0.09 – 0.61)

0.57 (0.36 – 0.85) 0.61 (0.39 – 0.92)

Intrinsic and extrinsic religiousness TJ Bouchard et al 95 Tabl e 6 Chi -squares and A IC i ndi ces deri ved from fi tti ng tw o model s to the M ZA and DZA covari ance matri ces for the Intri nsi c and Extri nsi c Rel i gi ousness Scal es

Scale IR ER

Full model (df = 4) χ2 P A IC 1.00 4.31

0.91 0.37

–6.99 –3.69

Environmental model (df = 5) χ2 P A IC 9.23 10.37

0.10 0.07

–0.77 0.37

Tabl e 7 Correl ati ons betw een the Intri nsi c Rel i gi ousness (IR) and Extri nsi c Rel i gi ousness (ER) scal es and the scal es of the Cal i forni a Psychol ogi cal Inventory (CPI) for the 195 parti ci pants w ho compl eted both i nstruments CPI scales Fol k scal es Domi nance (Do) Capaci ty for Status (Cs) Soci abi l i ty (Sy) Soci al Presence (Sp) Sel f-A cceptance (Sa) Independence (In) Empathy (Em) Responsi bi l i ty (Re) Soci al i zati on (So) Sel f-control (Sc) Good Impressi on (Gi ) Communal i ty (Cm) Wel l -bei ng (Wb) Tol erance (To) A chi evement vi a Conformance (A c) A chi evement vi a Independence (A i ) Intel l ectual Effi ci ency (Ie) Psychol ogi cal -mi ndedness (Py) Fl exi bi l i ty (Fx) Femi ni ni ty/ M ascul i ni ty (M F) Vector scal es V.1 Internal i ty V.2 Norm-Favori ng V.3 Sel f-Real i zati on Speci al Purpose Scal es Di cken Soci al Desi rabi l i ty (DSD) Di cken A cqui escence (DA C)

IR

ER

–0.07 –0.02 –0.04 –0.16 –0.15 –0.10 –0.05 0.20a 0.17 0.21a 0.18a 0.08 0.04 0.09 0.12 –0.11 –0.06 –0.05 –0.09 0.21a

–0.19a –0.11 –0.12 –0.14 –0.21a –0.20a –0.16 –0.10 0.06 0.01 0.08 –0.18 –0.06 –0.08 –0.07 –0.18 –0.22a –0.13 –0.14 0.09

0.11 0.15 0.03

0.13 0.12 –0.08

0.13 –0.10

0.00 –0.02

aP>0.01

The correl ati ons betw een IR and ER and RWA (0.47 and –0.10) repl i cate A l temeyer’s fi ndi ngs very cl osel y, tel l i ng us that the scal es behave si mi l arl y i n our adul t sampl e. Tradi ti onal i sm al so correl ates much hi gher w i th RWA than IR (0.71 vs 0.37), tel l i ng us that IR i s suffi ci entl y di sti nct from RWA to justi fy our treatment of i t as a di sti ncti ve measure. The correl ati ons betw een IR and ER and REL (0.67 and –0.22) tel l us that IR i s measuri ng somethi ng si mi l ar to REL, a measure w i th some construct val i di ty,32 w hi l st ER i s measuri ng somethi ng el se enti rel y. The assortati ve mati ng coeffi ci ents for IR (0.32) i s sl i ghtl y l ow er than one mi ght expect gi ven that Wal l er et al 32 report assortati ve mati ng coeffi ci ents of around 0.56 for a vari ety of measures of rel i gi ousness uti l i zi ng a somew hat l arger group of M ISTRA parti c-

Tabl e 8 Correl ati ons betw een the Intri nsi c Rel i gi ousness (IR) and Extri nsi c Rel i gi ousness (ER) scal es and the scal es of the M ul ti di mensi onal Personal i ty Questi onnai re (M PQ) for the 195 parti ci pants w ho compl eted both i nstruments MPQ Scales

IR

ER

Pri mary Scal es Wel l -Bei ng Soci al Potency A chi evement Soci al Cl oseness Stress Reacti on A l i enati on A ggressi on Control Harm Avoi dance Tradi ti onal i sm A bsorpti on

–0.03 –0.11 0.05 0.06 0.08 –0.04 –0.30a 0.17 0.09 0.37a 0.02

0.01 –0.09 –0.13 –0.02 0.05 0.06 –0.03 0.06 0.14 0.04 0.03

Hi gher Order Factors Posi ti ve Emoti onal i ty Negati ve Emoti onal i ty Constrai nt

–0.03 –0.01 0.32a

–0.07 0.05 0.10

a

p>0.01

i pants. The assortati ve mati ng coeffi ci ent of ER i s 0.13. The hi gher val ue for IR i s consi stent w i th the expectati on of assortment for i ntri nsi c rel i gi ousness, w hereas the l ow er val ue for ER i s consi stent w i th fi ndi ngs for personal i ty trai ts. Bouchard et al 45 report a mean spousal correl ati on of 0.22 for the 20 scal es of the CPI for 111 spouse pai rs i n the M ISTRA sampl e. A ssortati ve mati ng on the M PQ i s much l ess (0.13)46 i f Tradi ti onal i sm i s removed from the l i st. A ddi ti onal assortati ve mati ng data on IR and ER deri ved from l arger sampl es w oul d be hi ghl y desi rabl e. The fi ndi ng of some correl ati on betw een contact and separati on on ER i n the DZA sampl e i s a puzzl e because previ ous anal yses of M ISTRA data have been unabl e to document any correl ati on betw een contact and tw i n si mi l ari ty. Recal l al so that except for Ti me Together Pri or to Separati on these measures are rough approxi mati ons because the A UROS w as mai l ed to many of the tw i ns after the assessment at w hi ch the contact measures w ere determi ned. Contact betw een the tw i ns does not expl ai n any of the tw i n si mi l ari ty on the CPI scal es45 w hi ch w ere al l admi ni stered at assessment. That the i nfl uence of contact i s l argel y restri cted to ER for the DZA tw i ns i s consi stent w i th the l arger DZA i ntracl ass correl ati on (0.38) for these tw i ns rel ati ve to the M ZA correl ati on (0.24). The fai l ure to fi nd l ater contact effects gi ven the effect of very earl y contact i s, how ever, curi ous. We have another w ay to parti al l y check on the fi ndi ngs of no i nfl uence due to l ater contact. We are currentl y conducti ng a ten-year fol l ow -up on M ISTRA parti ci pants and have A UROS scores on ten addi ti onal pai rs of M ZA s and si x addi ti onal pai rs of DZA s. These tw i ns compl eted the A UROS onl y once, at the ti me of the fol l ow -up.

Intrinsic and extrinsic religiousness TJ Bouchard et al 96

When these tw i ns are added to the current sampl es, the M ZA and DZA i ntracl ass correl ati ons for IR are 0.35 and 0.17. The comparabl e correl ati ons for ER are 0.30 and 0.37. A ddi ng tw i n pai rs w i th addi ti onal contact hardl y changes the correl ati ons reported i n Tabl e 5. Gi ven these resul ts w e are i ncl i ned to i nterpret the M ZA data as a fai l ure to repl i cate the DZA resul ts and ascri be the fi ndi ngs to chance. The possi bi l i ty that ER (and perhaps other si mi l ar trai ts) i s easi l y i nfl uenced by contact shoul d be kept i n mi nd. The geneti c fi ndi ngs for E shoul d be consi dered tentati ve at best. The si mi l ari ty of the M ZA tw i ns on the M RE scal e of the FES (0.32) i s l arger than found previ ousl y (0.18) i n an earl i er and smal l er sampl e of M ISTRA tw i ns.47 Neverthel ess, our anal ysi s of the adopted vs bi ol ogi cal fami l y correl ati ons demonstrates that pl acement on thi s vari abl e can expl ai n onl y a tri vi al amount of the si mi l ari ty betw een our tw i ns as M RE i s onl y w eakl y correl ated w i th rel i gi ousness and pl acement on M RE i s onl y modest. Onl y envi ronmental vari abl es for w hi ch there i s both strong sel ecti ve pl acement and a strong i nfl uence on outcome can expl ai n tw i n si mi l ari ty on that outcome. On both counts M RE does not provi de an expl anati on for tw i n si mi l ari ty on IR and ER. Overal l , our fi ndi ngs that IR and ER both demonstrate a modest degree of geneti c i nfl uence confi rm previ ous behavi oral geneti c fi ndi ngs i n the domai n of atti tudes. The hypothesi s that IR w oul d show a hi gher heri tabi l i ty than ER i s not confi rmed. Data from a sampl e of tw i ns reared together coul d confi rm or refute these fi ndi ngs, and gatheri ng such data w oul d cl earl y be w orthw hi l e. The onl y addi ti onal fami l y data on IR and ER w e have been abl e to l ocate i s a parent–offspri ng correl ati on of 0.39 for IR reported by A l temeyer.29 He does not report the comparabl e correl ati on for E. Si nce the offspri ng i n hi s study are sti l l young thi s correl ati on coul d be consi stent w i th a number of hypotheses. A l temeyer argues that ‘the rel i gi on of authori tari ani sm i s al l of a pi ece, a sel f servi ng bel i ef system, acqui red i n chi l dhood and strongl y rei nforced thereafter, w hi ch requi res no external confi rmati on and w hi ch can probabl y survi ve i ntact i n the face of consi derabl e di sconfi rmati on’ (p 222). Thi s i s a cl assi c exampl e of i nterpreti ng a fami l i al correl ati on as evi dence of envi ronmental causati on. A parent–offspri ng correl ati on of 0.39 does not refute the hypothesi s of a moderate geneti c i nfl uence nor does i t support the hypothesi s of compl ete cul tural transmi ssi on put forth by the author. The same hypothesi s of parental reari ng as the determi nant of vari ance i n RWA , al so put forth by A l temeyer, has been show n to be i nconsi stent w i th a behavi oral geneti c anal ysi s of data from M Z and DZ tw i ns reared together and apart. RWA appears to

have a heri tabi l i ty i n the range of 0.50 and 0.64 and modest shared envi ronmental i nfl uence.48 A l temeyer,31 how ever, reports a parent–offspri ng correl ati on of 0.55 (n = 75) for a sampl e of adopti ve parent– offspri ng pai rs, a val ue that i s si gni fi cantl y hi gher than the bi ol ogi cal parent–offspri ng correl ati on. It w i l l be i nteresti ng to see i f thi s correl ati on can be repl i cated i n an ol der sampl e, as w e know that DZ correl ati ons for atti tudes drop dramati cal l y once the offspri ng l eave the nest.18 The DZA correl ati ons for IR and ER w ere not l ess than hal f the M ZA correl ati ons. Thus the hypothesi s of non-addi ti vi ty for rel i gi ousness measures suggested by previ ous research w i th thi s sampl e32,49 w as not confi rmed w i th a di fferent measure of the underl yi ng construct. Non-addi ti vi ty i s a form of i nteracti on, and i nteracti ons are notori ousl y di ffi cul t to repl i cate. A si mi l ar fai l ure to repl i cate nonaddi ti ve geneti c vari ance for scal es of the M ul ti di mensi onal Personal i ty Questi onnai re (M PQ) can be seen w hen the study of Tel l egen et al 50 i s compared w i th Fi nkel and M cGue.51 The fai l ure to fi nd substanti ve correl ati ons betw een IR and ER and the scal es of the CPI and M PQ i s strong evi dence for di scri mi nant val i di ty and sol i dl y confi rms the uni que nature of these measures of rel i gi ousness and the need to l ocate them i n the conceptual space of human i ndi vi dual di fferences. The modest posi ti ve correl ati ons betw een the soci al maturi ty cl uster of the CPI (Re, So, Sc) and IR are consi stent w i th reports that rel i gi ousness i s a protecti ve factor agai nst numerous debi l i tati ng psychol ogi cal condi ti ons and shoul d be i nterpreted as evi dence of construct val i di ty, as So and Sc are predi ctors of del i nquency.52 The correl ati ons betw een IR and the M PQ scal es of A ggressi on (–0.30) and Tradi ti onal i sm (0.37) shoul d be i nterpreted i n the same manner. Correl ati ons of thi s magni tude and i n thi s di recti on are just w hat one w oul d expect. The CPI scal es used i n thi s study have been show n to be moderatel y heri tabl e (mean h 2 = 0.46) usi ng M Z and DZ tw i ns reared apart and together.45 The l ack of si zeabl e correl ati on betw een the tw o sets of heri tabl e trai ts demonstrates that di fferent geneti c factors are probabl y at w ork i n each i nstance. The fai l ure of al most al l of the response styl e measures to correl ate w i th IR and ER cl earl y confi rms previ ous w ork suggesti ng that the IR and ER scal e scores do not si mpl y refl ect i nval i d response arti facts. Gi w as the onl y excepti on and the correl ati on w i th IR w as modest and not repl i cated by the Unl i kel y Vi rtues scal e of the M PQ. Chau et al 28 al so exami ned personal i ty correl ates of the IR and ER scal es and concl uded that ‘w hi l e the I–E scal e coul d be i mproved upon, i t assesses i mportant di mensi ons of i ndi vi dual di fferences and major cri ti ci sms of the

Intrinsic and extrinsic religiousness TJ Bouchard et al 97

scal e are unsupported or onl y margi nal l y supported i n the data presented herei n’ (p 400). Our fi ndi ngs and those of Chau et al support Donahue’s23 contenti on that IR i s a substanti ve construct of si gni fi cant soci al rel evance rather than si mpl y a refl ecti on of conformi ty, soci al l y desi rabl e respondi ng or acqui escent response styl e. Our resul ts, how ever, do not contri bute much i f anythi ng to the construct val i di ty of ER. Rel i gi ousness i s a posi ti ve predi ctor of physi cal heal th,12 a consi stent predi ctor of drug non-i nvol vement i n adul ts13,53 and adol escents,14 i s rel ated i n a compl ex manner to i ssues of prejudi ce and i ntol erance23,24,29 as w el l as mental heal th 54 and del i nquency.23,55 It i s cl earl y an i mportant construct, deservi ng of greater attenti on from soci al sci enti sts than i t has recei ved to date.

A ck now l edgements Thi s research has been supported by grants from The Seaver Insti tute, The Pi oneer Fund, The Uni versi ty of M i nnesota Graduate School , The Koch Chari tabl e Foundati on, The Spencer Foundati on, The Nati onal Sci ence Foundati on (BNS-7926654), and the Harcourt Brace Jovanovi ch Publ i shi ng Co.

Refer ences 1 Wi l son EO. On Human Nature. Harvard Uni versi ty Press: Cambri dge, 1978. 2 Goodenough WH. Evol uti on of the human capaci ty for bel i efs. Zygon 1993; 28: 5–27. 3 M aser JD, Gal l up GG Jr. Thei sm as a by product of natural sel ecti on. J Religion 1990; 70: 515–532. 4 Barkow JH, Cosmi des L, Tooby J. The Adapted Mind. Oxford Uni versi ty Press: Oxford, 1992. 5 Buss DM . Evolutionary Psychology: The New Science of the Mind. A l l yn Bacon: Boston, 1999. 6 James W. Varieties of Religious Experience. Longman: New York, 1902. 7 James W. The Principles of Psychology. Hol t: New York, 1890. 8 A l l port GW. Personality: A Psychological Interpretation. Hol t, Ri nehardt & Wi nston: New York, 1937. 9 Ehrl i chman H, Brody N. Personality Psychology: The Science of Individuality. Prenti ce Hal l : Upper Saddl e Ri ver, NJ, 1998. 10 M athew s G, Deary IJ. Personality Traits. Cambri dge Uni versi ty Press: Cambri dge, 1998. 11 Thomas DL, Carver C. Rel i gi on and adol escent soci al competence. In: Gul l ota TP, A dams GR, M ontemayor R (eds). Developing Social Competence in Adolescence. Sage: New bury Park, CA , 1990, pp 195–219. 12 Comstock GW, Partri dge KB. Church attendance and heal th. J Chronic Dis 1972; 25: 665–672. 13 Gorsuch RL. Psychol ogy of rel i gi on. In: Rosensw ei g M R, Porter LW (eds). Annu Rev Psychol 1988, pp 201–221.

14 Wal l ace JW, Wi l l i ams DR. Rel i gi on and adol escent heal thcompromi si ng behavi or. In: Schul enberg J, M aggs JL, Hurrel mann K (eds). Health Risks and Developmental Transitions during Adolescence. Cambri dge Uni versi ty Press: New York, 1997, pp 444–468. 15 M eyers SM . A n i nteracti ve model of rel i gi osi ty i nheri tance: The i mportance of fami l y context. Am Social Rev 1996; 61: 858–866. 16 Pl omi n R. Nature and nurture. Brooks/ Col e: Paci fi c Grove, Ca., 1990. 17 Cohen D. Stranger in the Nest. Wi l ey: New York, 1999. 18 Eaves L, M arti n N, Heath A , Schi eken R, M eyer J, Si l berg J, Neal e M , Corey L. A ge changes i n the causes of i ndi vi dual di fferences i n conservati vi sm. Behav Genet 1997; 27: 121–124. 19 Wi l son EO. Integrated sci ence and the comi ng century of the envi ronment. Science 1998; 279: 2048–2049. 20 D’Onofri o BM , Eaves LJ, M urrel l e L, M aes HH, Spi l ka B. Understandi ng bi ol ogi cal and soci al i nfl uences on rel i gi ous affi l i ati on, atti tudes and behavi ors: A behavi or–geneti c perspecti ve. J Personality (i n press). 21 Lykken DT. Stati sti cal si gni fi cance i n psychol ogi cal research. Psychol Bull 1968; 70: 151–159. 22 Wi l l i ams DR. The measurement of rel i gi on i n epi demi ol ogi c studi es. In: Levi n JS (ed.). Religion in Aging and Health: Theoretical Foundations and Methodological Frontiers. Sage: New bury Park, CA , 1994, pp 125–148. 23 Donahue M J. Intri nsi c and extri nsi c rel i gi ousness: Revi ew and meta-anal ysi s. J Pers Soc Psychol 1985; 48: 400–419. 24 A l l port GW, Ross JM . Personal rel i gi ous ori entati on and prejudi ce. J Pers Soc Psychol 1967; 5: 432–443. 25 Rokeach M . The Nature of Values. Free Press: New York, 1973. 26 Gorsuch RL, Venabl e GD. Devel opment of an ‘A ge Uni versal ’ I–E scal e. J Sci Study Religion 1983; 22: 181–187. 27 A l l port GW, Vernon PE, Li ndzey G. Manual for the Study of Values (3rd edn). Houghton M i ffl i n: Boston, 1960. 28 Chau LL, Johnson RC, Bow ers JK, Darvi l l TJ, Danko GP. Intri nsi c and extri nsi c rel i gi osi ty as rel ated to consci ence, adjustment, and al trui sm. Pers Individ Diff 1990; 11: 397–400. 29 A l temeyer B. Enemies of Freedom. Jossey-Bass: San Franci sco, 1988. 30 Hutchi nson GT, Patock-Peckham JA , Cheong JW, Nagoshi CT. Personal i ty predi ctors of rel i gi ous ori entati on among Protestant, Cathol i c, and non-rel i gi ous col l ege students. Pers Individ Diff 1998; 24: 145–151. 31 A l temeyer R. The Authoritarian Spector. Harvard Uni versi ty Press: Cambri dge, M A , 1996. 32 Wal l er NG, Kojeti n BA , Bouchard TJ Jr, Lykken DT, Tel l egen A . Geneti c and envi ronmental i nfl uences on rel i gi ous i nterests, atti tudes, and val ues: A study of tw i ns reared apart and together. Psychol Sci 1990; 1: 1–5. 33 Bouchard TJ Jr, Lykken DT, M cGue M , Segal NL, Tel l egen A . Sources of human psychol ogi cal di fferences: The M i nnesota study of tw i ns reared apart. Science 1990; 250: 223–228. 34 Segal NL. Entwined Lives: Twins and what they tell us about Human Behavior. Dutton: New York, 1999. 35 Lykken DT, M cGue M , Tel l egen A . Recrui tment bi as i n tw i n research: The rul e of tw o thi rds reconsi dered. Behav Genet 1987; 17: 343–362. 36 Gough HG. CPI Manual (3rd edn). Consul ti ng Psychol ogi sts Press: Pal o A l to CA , 1996. 37 Dahl strom GW. The grow th i n acceptance of the M M PI. Prof Psychol Res Pract 1992; 23: 345–348. 38 M oos RH, M oos BS. Family Environment Scale: Manual (2nd edn). Consul ti ng Psychol ogi sts Press: Pal o A l to, CA , 1986.

Intrinsic and extrinsic religiousness TJ Bouchard et al 98 39 M cGue M , Bouchard TJ Jr. A djustment of tw i n data for the effects of age and sex. Behav Genet 1984; 14: 325–343. 40 Tayl or HF. The IQ Game: A Methodological Inquiry into the Heredity Environment Controversy. Rutgers Uni versi ty Press: New Brunsw i ck, NJ, 1980. 41 Farber SL. Identical Twins Reared Apart: A Reanalysis. Basi c books: New York, 1981. 42 Neal e M C, Cardon LR (eds). Methodology for Genetic Studies of Twins and Families. Kl uw er A cademi c Publ i shers: Dordrecht, 1992. 43 Neal e M C. Mx: Statistical Modeling. Department of Human Geneti cs, Box 3 M CV: Ri chmond, VA , 1995. 44 A kai ke H. Factor anal ysi s and A IC. Psychometrics 1987; 52: 317–332. 45 Bouchard TJ Jr, M cGue M , Hur Y-M , Horn JM . A geneti c and envi ronmental anal ysi s of the Cal i forni a Psychol ogi cal Inventory usi ng adul t tw i ns reared apart and together. Eur J Pers 1998; 12: 307–320. 46 Lykken DT, Tel l egen A . Is human mati ng adventi ti ous or the resul t of l aw ful choi ce? A tw i n study of mate sel ecti on. J Pers Soc Psychol 1993; 65: 56–68. 47 Hur Y-M , Bouchard TJ Jr. Geneti c i nfl uence on percepti ons of chi l dhood fami l y envi ronment: A reared apart tw i ns study. Chil Dev 1995; 66: 330–345.

48 M cCourt K, Bouchard TJ Jr, Lykken DT, Tel l egen A , Keyes M . A uthori tari ani sm revi si ted: Geneti c and envi ronmental i nfl uence exami ned i n tw i ns reared apart and together. Pers Individ Diff (i n press). 49 Di Lal l a DL, Carey G, Gottesman II, Bouchard TJ Jr. Personal i ty i ndi cators of psychopathol ogy vi a M M PI i n tw i ns reared apart. J Abnorm Psychol 1996; 105: 491–499. 50 Tel l egen A , Lykken DT, Bouchard TJ Jr, Wi l cox KJ, Segal NL, Ri ch S. Personal i ty si mi l ari ty i n tw i ns reared apart and together. J Pers Soc Psychol 1988; 54: 1031–1039. 51 Fi nkel D, M cGue M . Sex di fferences and nonaddi ti vi ty i n heri tabi l i ty of the mul ti di mensi onal Personal i ty Questi onnai re scal es. J Pers Soc Psychol 1997; 72: 929–938. 52 Gough HG. A pprai sal of soci al maturi ty by means of the CPI. J Abnorm Psychol 1966; 71: 189–195. 53 Gorsuch RL. A n i nteracti ve, mul ti pl e model approach to i l l i ci t drug use. In: Letti eri DJ (ed.). Theories of Drug Use. Nati onal Insti tute of Drug A buse: Washi ngton, DC, 1980. 54 Bergi n A E. Rel i gi osi ty and mental heal th: A cri ti cal reeval uati on and meta-anal ysi s. Prof Psychol Res Pract 1983; 14: 170–184. 55 Stark R, Kent L, Doyl e DP. Rel i gi on and del i nquency: The ecol ogy of a l ost rel ati onshi p. J Res Crime Delinquency 1982; 19: 4–24.