The Splitting Index:

Constmction of a Scale Measuring the

Downloaded by [Central Michigan University] at 17:14 04 January 2015

Defense Mechanism of Splitting

John R . Gould. Norman 34. Prentice. and

Ricardo C. AinsIie
V~riveasrtjof Texas at Austm

The Spkirring Index (SI), a self-report scale based on the wrirings of Kernberg
(e.g., 1976) on seif and object rcpreselatations and the defense mechanism of
spiitrrng. was construcxd. After deveioprnect over the course of 6 pilot stxiies,
the SH was validated through 2 fur:ber studies hcior analyses revealed z
24-iiem scale w:th :hree &item sohscales, mezsuting the spiirting of self,
%miIy: and others' images The S I znb its slihscaies were denmstrared to be
internally consistent and stabie over a 4-week period. Convergent validity was
snpporred by significant correiations wirh measures of borderijne and marclssis-
tic personzlity disorcIess, self-imagc s~abiiity,self-esteem. depression, arid neg-
otiz:e affect:vity Disctiminmt vaiidity waa ciencnstra:ed by near-zero
correia:ior,s w i i h two measures of cognitive compiexiry. Cnnrrary ti: predjc-
tions, the SI was sigr?ifxanlly correlated with me Dogmatism Scale (Rckeach.
1969j. a third measure of cogci~vecorxplexlty. Research and ciinica! applica-
tions oi'the St. zre discussed.

The defense mechanism of splittrng has been widely associated with patients
who are fcnctioning at the low end of the psychopaihology s p e c t r ~ m
(Kernberg. 1976). AIhough spEitting 1s especially associ;ated with :he diag-
nosis of borderline perssnaiity disorder, Kernberp has e r n ~ h a s i r e dthat this
mechanism of defense is best considered reflective of a particular kind of
ego o r g a n i z a t ~ o En
~ theorizing about the o r i g i ~ ss f the splitting defense,
Kernberg draws from Piagetian (e.g., Pmget 1937/1954) concepts of early
schema developaenr, hypothesizing that children initiajiy organize experi-
ence into consteikaticens of good and bad on the basis of their common
affecf ve kalence. Kernberg (19776) proposed that, in normal deve'Iopmene.
good and bad self s a d object (ozherl representatioas gradually became i n k -
gr3ted into whole self and object representatrons as cognitive maturation
 proceeds in the third year i?f life, and splitting is gradtraily replazed b:i
repression as the centrai mechanism of defense.
However, when there has been ri predorninancc o i h a t r a t i n g experiences
e,irip In deveiopment. there is an interference In the normal integratior: of
good 2nd bad self-objeci representations into more reaiistic perceptior:~cf
self 2nd others. Kernberp i1955) hypothesized that urider these circt;!n-
stances, the norma! develo~merataiprocess i n which spilxinp gi\.es w a y to
integrated represenlaricns becomes disrupceci. Instead, splitring b e c o m e
enlrencheii and utiiized defensively in an effort tt: maintain intrapsychlc
Downloaded by [Central Michigan University] at 17:14 04 January 2015

stabiiity and cnheslon. It is a1 this jisnclure that spiit:inp becomes the M i -

:nark defense that many authors haFe observed jn b a r d d i n e l n a x i s s i r i c .
arnd other severe personalit\i disorders I G i o ~ a c c h i n i . 19P3; Guildersoil,
1385: Kernberg. 1975; hleissner, 1984. i988 i.
Ciinicnliy. splitting is distinguished by a terdency to see oneself cv others
as 311 good or all bad. Indh~iduniswho frusnate. disappoict. or i: some orher
manner interfere u.irh one's perceived needs ma become carepori~ally
vielved as bad and evii. whereas individuals who gratify or s , h o :ire p t e n t l a i
sources of needed responses tend to be \-ien,ed unidimensionaliy as nil good.
One of the characteris~icfearures of splitrlng as a defense mechanism is :he
instability i t creates. For exarcpie. the same indiviiiuai is ofter: subject to
widely shifting perceptions depending on whether rhat person. is seen as
. ..
gratifying or f r ~ s x a i i n gat any particliiar mornenc. There ;ire i~mria:c+cil!a-
tiuns in the experience cf self. Thus. this kind of' a defeniri1.e srrwture reiidii!,
lead. :o chaotic and unstable inaerpersocai relationship patterns. identity
diffus>on,.and mood swings. I n the t h e r a p e u t i c s e t t i n g . r c l i a n ~ eor: the splir--
ting mechacism of defense presents an impediment to the :herapeurl;: y o -
ccss and requires ccxsiseent interpretation tc, overcome its dele!erjous effects
on treatment outciprne ( Kernberg. 1975).
Given the extent to which sptittinp i s enconntered i n i.lir?.icz!,~r'orkwit5
sei:e-e:y disturbed patients, :be d e ~ e ? o p m e n of
t a measure of this defense
mechanisn: is potentially of great \.slue. Research on the spl~ttingdefense
mechanisn: has been successfu1iy conducred u.sing the Rorschach Test (e.g.:
t e r n e r 22 Lerxx-, I 9 X ) and clinical lnterviews (2.g.. Perry. i983!.F f o ~
e\.e:,.
there have been few empirical attempts ro d e w l o p a seif-report question-
naire of spiitting. Irr what was perhaps the first s!;sterna:,c nt!enpr. Gerson
! 1984) constructed a 14-irem seif-report measure of deferrsi\,e splitring based
cin the v;ri:ir:gs of both Kernberg (1955,) and Kohu: (14-1'1. ReimDilit>
analyses revealed a standardized iten-: alpha coeffizien: of .72 and a test-re-
test reliability correLation of .84 fp < "001). A factor nnalysis yielded a
rhree-factor scll:tion. inzluding z major factor of 2ene:d s p i i x k p ;conrain-
ir:g I0 items). Glassmar: (1436; failed LO replicate rhe factor strucrure of this
measrrre and cjuestioned th:: psychometric value of seve:al of :he pIi:ting
items. Many have noted that Kohu: and Kernberg conceive of spli:ting
prc!cesses f r o m widel): different p s r s p e c t i ~ e s:e.p.. Glassman. 1986:
Mer~t.her_r,i983; Marmar & EIoro\virz. 1986;. Consequei~r!y.[he conxprdai
 incongrulry in Gerson9s ( I 984) approach may havc accounted for the failure
of her s c a k to cross-valj6ate. For our research; we selected Kernberg's
::1975. 1976. 1984) conceptualization of splitting because he has elaborated
this ~ e c h a n i s mmost extensively in his theories,
Using s traditiomi psychoanalytic perspective. Bond ( 199i ) constructed
the Defense Style Questionnaire ( D S Q ro measure various mechanisms of
defense, induding a ?-item ciuster measuring the spfitting defense mecha-
nism. A factor andysis of the defense mechanism cirnsters yieided four
factors, called i l e f e ~ s esryles by the authors (Bond. Garcine:; Christian. &:
Sigal, 19831. One defense style was labeled image-tiisror-ring deftmrs and
Downloaded by [Central Michigan University] at 17:14 04 January 2015

conrained the spiitting defense mechanism cluster as well as other closely

related defense mechanisms. A n d r e w , Pollock. and Ste\n!arr i1989) were
u m b l e io replica~ethe four-factor solution of the DSQ with either a
~sychirttricpatient sample or a nonpatient sample.
Bond et al. (1989:; were unable to provide vniidatinp evidence for the
DSQ with the Defense Mechanism Rating Scale (DhlHRS; Perry, 1967,based
ctn ciinicai nter\:iews. The Image-disiorting defenses of the DSQ did not
correlate significan~!y with the same-named grorip of defenses of the DMRS.
Sirnliarly. the splitring defense ciuster cf the DSQ did not correlate signifi-
can~E:, w:rh the ,vpiirting qfw(f-imagr: or the sp!iriing qf o t k ~ r s 'i m i g r s
dederrses of the DMRS. Bond and Vaillant i1986) faiied to accurately predict
rnnjor diagnostic categories of patients, including a category of severe per-
sonality disorders: using the four defense styles of the DSQ. Similarly. Bond
11990: did nor find a higher usage of the image-distsriicg defeense style in
borderline personality disordered patienrs ihar, ic other patien: groups. Over-
all. research has not produced satisfactory vaiidity evideaee for :he image-.
d i s t o r t i n defense style of the DSQ.
The pcrpose of our research was to develop a seif-report measure. of
sp.Ei:ting with satisfactory psychometric properties. Such a measrrre u - o d d
provide a cost-efficient and corrvenent instrument for empirical investiga-
tions of psychoanalytic object-relations theories of splitting as weil as of the
borderline and i?arcissiseic personality disorders. The Splitting Index (SI!
wouHd also be useful as an initial screening device for identifying severe
charactemlogical pathology and predicting potential treatment difficulties.

Parficipanfs and procedure. Based on the theoretical wrrirnps of

Kernberg (I945 1976). a number of Items were d e i r i e d to measure the
ciefense rnechanim of spi:rtmg Item\ were also c o h c ~ t e dfrom clm:aI
psychologicti and psychistrists ~n the communrry b j askacg :hem to write
 s m e m e n t s correspo:~dinp to a brief descriptioc of spiitrinp. 111 most i n -
stmces. these sciicired statements were sim,iified to enrcre unifarmr- and
dlrecrness in the i t e m . Aprosirnraiejy one half of the items \ v e x ix::ended to
measure the presence of the spiitclng defezse mechricism. and the rtmzinder
were intended to measure the absence cf spiittjnl;.
The SI ,was deveBoped over the course cf six pilot stndies. Successi~-e
\.ersjons of the SI were administered to a total o f i .;8 1 andergr3dua:e
srudents (56% mez'i from the L:niversiq of Texas a: A i i z t i n u.hn p t i c i p a w d
to firlGIi introduc:ory psychoiopq- course requirements. These students were
Downloaded by [Central Michigan University] at 17:14 04 January 2015

predominantly freshmen. White. and from xiddje-class bxkgrcsands. The

pE!ot studies had a median number ef 226 students irznge = -7 to 5 6 6 ; .
Students were asked to rate :heir ngreement or disagreement u,ith tach i:em
or? z Likert-type scaie razging from I (sr!nrrgi.< ~r!i.rag!-c.~~, to 5 cc~ro)rai'i.
. .
,. .
r ~ ~ g w e , , For
) . :he purposes of anaIy.sis. the ratings :or ;:ems intended rn mea-
slrre rhe absence of the splitting defense rnechanisx were ~s\.erseii.so that 2
5 indicated a high degree of spiltting for each item The score for each
s~udenrwas the aierage rating nf a11 the items. resdting I n ;1 possib!e range
of I to 5 .

ST i t e x s were retained o: discarded based on re!iabi!i:!. ~!r:,Iysrs!Cronb;ich.

195; 1. with new items constructed and added to the .;:air foIlois.ing e a ~ h
p i i ~ study.
i -4 rnirrimurx: c.f 5 0 splitting items were inttiaII1- :nclrtdt.d i n each
version cf the SI. The skxrh pilot study resuited in a 38-itern SI ( X f = 2.-K?. SD
= .49. a = .8S:. Male ( M= 2 . 4 4 and female i M = 2.331 partiitpants' scores
did n o t differ signit'icantIy on this version ofrhe SI. ttZ7-l-i = i .76. effect size
(Glass, i 947) = .22.

STUDY 3

Parficipnts anbproeecdures. Because the s p l i t t i r : ~defense mechri-

nism is commonly associated with borderline an6 narcisistic persnr,r:,Ii::s
disorders r e , g . Kernberg. !975: Meissner. 1984): measures nf these diapnm-
 tic categories were adrnin~steredand predicted to csrreiaie positr~el!, with
the SI. One of these measures was the Borderlrne Syndrome Index (BSI.
Conte, PIurchik, Karasu, & Jerrett, 1980). a 52-item questronnaire desrpned
to measure the borderline personalirq disorder Also mcIuded was the
Schizotypal-Borderline scale (STB, Clandge & Broks. 1984). w h ~ c hI \ com-
posed of I8 statements based oa the Dicgno~ficand Si~trstical.&fcime?iof
Mental Disorder: (3rd ed [DSM-f1C; American Psychmric Assoc~atior?
[APA]. 1980) criteria for borderhne personality Jiscrder. In addrtion, the
h'arciss~sticPersonahty Disorder MMPl Scale (YPD: Ashby. Lee. & Duke.
Downloaded by [Central Michigan University] at 17:14 04 January 2015

i 979) was sdrninistered The NPD consists of 19 staternencs from the hlinne-
w t a h3ealtzpha~rcPerscmahty Inverrtory (Hathaway & McKrnley. 1 9 4 3 ) that
reliably differentiated narcissistic personality dnsorde: patients from a con-
trol gnoup. For each of these scales. rterns are ansnereci rn a d ~ h o t o m o u s
format (yes-no or true-false)
Furthermore. borderline and other severe pe-sonalit'; drsorderi that ap-
pear to use the sptrtting mechanasm of defense are often associated with
~ncreasedlevels of depress~onand anhedonia ie 2 . Gunderwn & Srnger,
1975. Kernberg, 1984:. Therefore. the SI was pred:cted to correlate posi-
trvely with the Beck Depression Inventory tBDI; Beck & Steer. i9e4),a
widely used meascre of depresvon. The 3 Itens on this scale are rated on a
4-point scale (ranging from 0 t s 3:. produc~nga p 0 s ~ i b krange of score5
from O to 61 The SI was a h predictea to carrelate povtive'lj wi:h the
Xepstive PLffectw~tyscale (%A, Tekgen. 1 9 8 2 . and io correlate negatlvelq
w;ih the Posita~eAffecrlv~tyscale (FA. Tdlegen,. 1962). TeDegen reported
that pa.itt:ve and negative affectrvit) are tvco general factors of the Multidn-
mens~onalPersanalrty Questionnaire (RIPQ, TeiIegen. I51$21 [hat "seem tcr
index the drengrh cf the rndlwdual-s disposition to experience. respectrvely,
pleaswe snd p a n , reward and punishment. self-enhancement and self-m-
perilment. and to behave and think i n wags that are conducnke to these
exprrknces" (p. 3) FolEowing the research by Watson a ~ cniIeagues d !e.g .
ViTatson. 19SS), posit~Teaffectiwty was assessed in thi5 research by the
I l -ttem Wellbeing scaie of the MPQ and nega':ive affect~vrtywas measwed
by the 14-item S ~ r e Rextnon
s~ scaie of the MPQ The items on these scales
are answered In a true-false formit.
Discriminant validity was also examined. It was hypsaheslzed that ~;plrt-
trng is a mechanism employed b j the ego :Q reduce mxietj t\> separztxg the
aggresslveiy determmed self and object represeretst~onsfrom those that are
deterrimed by posit~veor pleasant experiences. Therefore. :he SI s9ouId not
be associated with the cogn:rrue compEexrty-simplicity dinension that refers
to the degree of attitude differentration exhihted by an :ndav~dualiBzer:,
1355) Three commonly used scales of cagnitibe complexity were uwd ia
Studj I . The Ia:oierance of Ambruity Scale ilAl was created by Budner
(1962) to measure "the tendency to gercs:ve ( i e . ~nterpret)ambigums
srtuations as sources of threat" (F 2 9 ) The Ih is a i 6 - ~ t e mscale with half the
 items keyed in the reversed directior:. The California f k a i e measuring
autborirxianism (Adorno, FrenkeI-Bnacswik, Levinson. 6 Sanford. 1950: is
campr)sed of 30 statements and each itexi is positi\.ei> k q e d . Rokeach's
11960) D o g n a ~ i s mk a l e [Form E j was cseci as the fin,?.]: ~ x a s x eilf ci7gni-
:jve coxplexiry. The Dogmotism Scale is composed of -13 statements. e x h
iniiicati7i.e of dognztism. Each of the cognirl\.e comp'iesity meascres :+ere
rsred cn a 5-point Likert-iype scale. Scores were averaged resrrlticg ir: a
possible range of i to 5 .
. ,
The SI and the construct validity n-ieasures were n:tm:nrsrered ro 811
undergraduie students from [he V2i\.ersity of Texas at Ausrir: i5C.65 men)
Downloaded by [Central Michigan University] at 17:14 04 January 2015

who participa~edto fulfiIf introductnr> g s : \ c b e i ~ ~cixErae

y rc.quIrement>.
Participacrs rated their responses to the SI items 01: the s a n e 5 - p o ~ nLikert-
t
rype scaie used in the piion studies, KIne new. items v.we added t a .:he items
s.;rvi\-ing the sixth pilor study groduc:ng a toral of17 SI items.

Irtterna! c~nsistency. RetiabiIIty analysis of the hii SI produced sn

aipha coefrlcjenr of .02. Six i ~ e m were
s r e m c ~ e d which
. maximized nipha at ,411.
The rescltinp 4 i -item ST had a mean of 3.32 and a standard deviation of .63.

Scaie structure. The 41-iten: SI was scbjec:ed to fac':or analyt~cpro-

cedures u s n g principal axis factoring with zn obhque rotation. The resnlts
nf this procedure reyealed three major factors within the SI q'ith 37 i:erns
Ioadinp above .3G on at Least one af the factors Foxteer: items reizteci :o the
splittjng of the self-image ioaded OR the principal faitor <hereafter referreil
ro as rhe Self factor) with an? eigen\ d u e of i 0.27 and accounti2g for 25.1 C+
of the \.ariance. Fac:or II consisted of 17 items relsted to ihr splitting of
f ~ r n i l ymembers' : m a p (hereafter referred to as the Fami!! factor! with 3r:
eipecvajne of 3.20. Elever: itenas pertairling to the spiirting of'others' !e.g..
.'$-:
,:rends." "pe:zp~e"~images loaded on Factor III ihereafrer referred to as the
Others factor) with an eigenvalilf: cf 1.72.
A: this pin:. Items were deleted based on [he results of the factor m a l y i s
to achieve the f01iov1.in.g goais: (a) ro :educe the total ncr:rher of items. (b) to
accentuate the distinc:ions among the factors. [c'l lo retain an ecjlini numbe:
cf items per subscale: and (ci) to remin a baiance of forv.3 r d and rei,ersed
i t e n s in the total scale. This process resuited in a 24-item scale fe:.turing :2
reversed irems and 8 items per fx:or. This revised scale ~ v z sthen resnbmjr-
tee lo the factor analytic procedures and :he three factilrs were replicated.
The results of Lqe final factor analysis are. presecred in Tabie I . Each factor
was rignificantlj, correlated (p < .001! with the other factors: S e l f and
Family? r = .29; Self and Others, r = "38:Family and Others, r = .42.The
reliabiiity analysis of the 24-irem SI produced an alpha coefficient of .9C! (?ii
 TABLE 1
Factor Loadings on the 24-Item SpIitting Index id?i&ining Principal Axis
Factoring With Obls~ueRotatson

Factor Item I 11 m
Self i2 ""8 "03 - .M
16 " l
99
I .02 "O?
5 .76 .@ .OI
:4 .74 .I0 .w
Downloaded by [Central Michigan University] at 17:14 04 January 2015

It .71 .M .02
24 .67 - .83 ,Ol
-#
.5Q - .02 .Q7
B .54 - .Mi .02
Family 17 "03 .?4 "07
23 -74 - .Ot
22 .1I "7% - 07
B - .09 .6B .05
4 - .11 .66 .OR
2 - .04 .62 - .55
8 .?@ .5? - .Ol
6 .QS A? .@
I3 - .06 .CO .7l
19 .oa - .02 70
21 - .0I .ii .67
3 - .03 - .Q4 .62
20 .I4 .063 .59
I E; I! l4 "58
I0 .I3 - 15 55
B5 - .Of "09 .48
Kofe. h' = 841. Factor I (eigecvalue = 6.9$)is labeled Spllrting of the Self Image, Factor
EI (eigerevakne = 2.46) 1s iabeled Splatcmg of Family Images, and Factor III (eigenvaine = 1.55)
is labeiec! §plattang of Others' Images. See Table 2 for list af items.

= 7.%, SD = 651. The SI Items and their rtern-total correlaucpm are presented
IE Table 2.
The alpha coefflcients fcsr each of the subscales were also high the Self
factor. a = 89 ( M = 2.93, SD = .98); the Family factor, a = .r(5 rM = 1 73. SD =
. V I .and the Others Sac:m, ex = 3 4 :M = 2 19. SD = .76>.In addktbn. each af
the subscaies was s~gn~flcantIy cp < OGI) correlated with one mother and
with the fell scale Thjs rnformat~onis presented In the top half at'Ta9le 3.

Gender d~fferences. The St dad not reliably dlfferentiare male iM =

2 26) and female ( M = 2.301 participants. ~ ( 8 2 - 4=) - 90, effect size = - 07.
Howeber. some gender differences emerged w i t h i n the subscales For the
Self tsctor. women (41 = 3.03) sccred srgnifrcantly hgimer than men ( M =
2 83), ~ ( 8 1 4 =) 3.04, p < . O i , effect size = .21. Women IM = I 76) also
 TABLE 2
Item-Total Correlations for the 24-Item SplB!:ing lnasx

Item
i. I feel differen: about myself when I with different geapie. (S)
(-) 2. My mother has faults, Su: I have never doubted her h e For me. (F!
I - i 3. Being able to keep friends is one of my strong polnts. (0)
i - i 4. My p a r e m always took a r e of m y neecis. F)
5 . My feelings about myself skfi dramatically. (S)
6 . Ir is impossible LO love my parents dl the time. (FJ
7 . Tna different parts of my personality are dfficillr to put
Downloaded by [Central Michigan University] at 17:14 04 January 2015

copether. is)
8. My feeiings asout my mother change from dzy to day. (F)
( - 1 9. My parents did the best they could for me. CF)
10. 1 hinve doub:s about my closes; friends. (0)
11. Somerimes I am not sure who I am. (S)
12. My Feeidngs abou: myseif are very powerfui, but they c m change
from one moment to the next. (S)
r - ) 13. MI friendsbps are airnost aiways satlsfyng (0)
( - ) I& Mj feekgs about nnyseif do not change easnl) {S,
6 - r IS I hake had man? long-iast~agfnendshpi, ( 0 )
15 1 sonretmes feel "pulied apart" bu my feehngs about rnyseif (SI
r - ) Is Mia? relat~oashpw t h rn) fami!) is solid (F)
r - 8 18. M j fedings toward those close to me rernam c ~ n s t a n t(0)
t - l 19. L have aways beer, aware that my dose friends red& cared
fo- me. (0)
I - ) 26. M) opinions of my frrends rarei? change. (01
i - ) Si I almost alwsys feel good about those close to me. (0)
22 I have extremely mixed feelangs about my mother. (T)
23. My family was often hu:tfuL to me. IF)
24 Who I am depends on haw 1 run feehng. (S)

Tutai scaie (alpha):

Note. ,V = MI S d e key ( - 1 = atem score reversed: (S) = Splmhg of Self Images
subscale; gTj = Splitting of Famiiy lmapes subscale: (0: = Sp!irtmg of Others' Images
subscale.

TABLE 3
Splitting Index (Slj Subscale CarreBation Matrix

h o f e AL conelatioms are significant @ < Mi).Correiaaons above the d~agonalsare

based on Stud) I (A' = 841) and correlations beiom the diagonals are based ox Stud\ 2
( N = 79) Reliabrlip esamates (alpha) are merteci m the diagonals. and were obtained d u r q
Stub) I
 422 GOUtD, PREKTICE. AiNSLfE

scored higher than men ( M = 1.70) on ihe Farn~lqfactor, but chic dlfter-
erece djd no( reach s r p i f l c a c c e , l(824) = : 32. effect srze = - 08.
Meanwhile. men Ih.f = 2 . 2 4 ) scored s ~ g n i f r c ~ n t lhyg h e : than women
!Ad = 2.10) on the Others factor, ~ ( 8 2 4 )= 2 60, p < 0 5 , effect slze =
i 8 Thus, the u o n u x in thrs \anr,Ie splnt the self-rmbge mclre than mer, Jici,
whereas the men kphl others' images more than women d ~ dOnly the resuIts
of the splitting of the seif-image were consistent with the reported higher
prebaler~ceof the borderline persocdity disorder dragnos]\ among women
i,APA. i 994)
Downloaded by [Central Michigan University] at 17:14 04 January 2015

Construct v~3kftfy The S1 and each of 11s subscalec correIaled sigs,h-

canCy ( p < c>OI) with both rsleasares of the borde:line personalrty drsvrder
(5% and STB) and the measure of the KPD Exaxrnatlon cxf Table 4 rebears
that these correlarron\ are consis:entIy k g h and mdrcates that the SI is
\ t r o ~ g l qrelated to these drtorder-5 ' h e ST 2nd ~ t s ~ ~ b s c a l ae cl w hsd srpnifl-
cant : p < O G I ) though <lightly lower cerrelaticns wi:h the BDH, XA. and PA
scale., Thus, t'ze SI appesrs to be assoc~atelrw,th i i e p r e s s ~ ~as n well as
negative 2nd pnsltlue affectlvrt;). AIE of the ~onwergent a!:drrq cnrrelat~onc
are presenxd i n Table 4
The dlscrrmnant walidrty anaiysis of Stud> i reiizred measures of cogni-
t r j e complexit? to the SE These result\ are pre\ented in T h l e 5 . Examrna-
tlon of Table 5 reveals the S.1 and les subscaies had low co:relations w t h the
IA scale and the Californra F S c d e Althcragh seteral of these correlations
reached sta:~siiaal srgnrfrcance. they arc rrniformly i e u (rxigc: = - 03 to
- 14) and appear to he artrfacts of the k r g e sample size rather than b a n g
~ndrcat:veof meanrngful reIatmcrns between the Sf end these measures. The
ccrreEat~onsb b ~ ~ l trhe
h Dogmstrsm Scaie were somewhat hrgher The SI and
the Self and Orhers subscaIes had srgnrf~cant( p c .OO!) corre~ancneswrth
Dogmatrsrn. indicating some degree of assoclat~onbetween the ST 2nd thrs
meacure of cognitike complexity

TABLE 4
Convergent Validity CorreBaticana for Study 1
Scde BSH STB NPD Be>i .VA P.4
Self -67 .BI .58 .53 "56 - .39
F amlIy "48 33 .37 .JO 20 -22
Ochers .49 .41 .47 .36 .39 - .33

Splitting Index .57 .60 .52 53 .59 - .4I

Syndrome Index: STB = Sclnieotyp&LBorder%in9scale; NPD -

Note. N = 841. MI correiations are significant @ < .MI). Scaie key: BSI = BorderBine
Narclssistnc Persondnty
Diso~de:MMPI Scale; BDY = Beck Depression Ivenhory; NA = Negative Affectiuky; PA =
Positive Affectivay.
 TABLE 5
Discriminarlt Validity Correlations for Stddy t
SCQL IA f -SC BOG
Se!f - P? . IO** 32-8
Family - .03 - :44$" M
Others - .08* - .Was 128'"

Splitting Index - .OT* - .& .--

-t**+

,Vote. ,V = 841. Scale kel: : A = 1n:clerancr of h S i g ~ l c y Scale; F-SC = Cakfornra F

Scde; DOG = Dogrnat~sn:Scaie
Downloaded by [Central Michigan University] at 17:14 04 January 2015

* p < .05. * * p < "01. *+*p < .MI.

STUDY 2

'The temporal stabiiiiy of the SH v.as in~estigatedic Study 2 . This study also
;ittempred to replicate the i.oi:vergent validity reziilts in\.cll\.iilp the r:-iezsiires
ai^ the borderline persona!ity disorder ilsed in Study I . Fr:,r:hrr ;on).erprcr
~ a l i d evidence
r~ A,as soughr through the inciusinn elf ;Idditic?ml. measures
~ r e l i i c t e dto have significant positive correiations with rhe Sl.

Pa:?ici~antsand procedures,. Participan~swere solicired from iniro-

. .
ductory psychology classes zi the University of T e x a a! Auhtic and partlcj-
pated :a> f u i i l l l cuvrsc reqr;rrrments. T\\:o ques:iannaire pscksts conrninirig
the SI were administered to separate samgles daring :he s m e semester.
. .
Thirty-one particrpants ( 1 3 men and i 8 ui.omen) receiveii the szme questinn-
m i r e pacicri a second time foliowing a 2--week i a t e r ~ a fcli conduct a tesr-re-
rest analysis of the SI. Because relati\!eEy few participants signed up to
partrcrpate i n the test-retest study. addktiocal partkipan% u e r e soiici':eci
from the same parricipant pool to complex one of :he questior.nair: p c k e t s
iil: c n e o c c a s o n cn1.r-. Therefore. a total of 79 participants i39 men and li
. .
women, with I irsdividiiai who failed to indicate gender! partjc:pa:ed in
St:rdy 2 and completed m e of :he two questionnaire packets Including the SI
on r:? leas: one occasion.
One questiomaire packet in Study 2 was administex3 to 42 participants
;:nci contained three neasures of the borderline personaiiiy disorder. These
included the BSI and STB measures used irs Study i . The third :r:easure of
the borderline persaxdity disorder was the Millon C!ini:aI hlaltiaxial I~-i.rn-
tar!-11 Borderline Personrilir~Scale (MCMI-TI: !v"/lI!io:!, i38-1. This scale
contaics f 1 prototype iterns. 14 sezondary :rexs: and 32 tertiary irenrs!
pmi.'diacing n total scnie of 62 items in a trne-false ibrmzt The pro!ot)pe
Items are weighted 3, points each, the secondary items zr-e \;.eiphred 2 p o i ~ r s
each., and the tertiary items are weighted h peint each, ibr a rota1 of 103
possibie points.
 The other questionnsrre packet wac, administered to 3' partlcipmts anc;
contained three sozra8 psychological mezsures concernlcg the <eif-can-
c e p as well as a measure of image-dlstortrng defenses Becauie \pirtting
of :he self -image produces idecrrty d i i f w o n (Kernberg, 1975 1. st was
predlcre3 thar the SI w o ~ ~ be i d raegart\ely a s s o c ~ a t e dwith rhe Rcasenberg
t 1979) Sta3:lity of Self scale (RSS), a 5-item Guttman scake meacuricg
self-c-oncep? stahrlity The item\ of the RSS are cjuektlnns or statements
f o i l w e d b? either two or four res=pon\e aiternatlves that are ranked In
terms or' <tatril~r)nf self-concep e a c h iten; ic x o r e ~ir any Pur the
lowest sxbjil:> of self-csncept alrernatrve rs reEected Tkus, the scalc has
Downloaded by [Central Michigan University] at 17:14 04 January 2015

a pc15~'bIer m g e of Cp to 5 *.?% higher s c o r e s rrrdizatlcg i ~ i g h c rst~br!:fy of

the wIf-cancept.
Hn add~tjon.Iow self-es:eem hits twen assocrateb w:th \pI~t~:r?g (Kernberg.
1995, and instd51i:rq c~fthe self-cerrcep: (Rosenberg, 1979). Thus, it was
predrcteh thar the SI woulci be nega:,\eIpi asso:iared w:th rwc measures of
\elf-esteem the Rosenherp i 1979: Self-Esteex \<ale IRSE: dnd the Texas
Social Behawor Hn~errtory(TSBI. E-lelmreiih Stzpp 6 Er\ in 19741 The
RSE is i0-ntem Guttman s c d e that iombrnes ilenri -ewi:r=g i n 2 poisibje
range of 0 to 6 H a 3 the rtemi rrre ke! ed In the relersed Sirectim. The TSBI
I S a 32-item ~ c d edes~gned10 metcure ind,b:duai perceponc: of social
competence 2nd \elf-e\tee= Each statement i > rated on a - p o r n t Llkert-r?pe
s c d e kilow i n g a possible range of CE f~ i2S Ten I t e m are keyed :n the
re\ler\ed drrectron. Frnally, rhe DSQ'c image-d~\rortrngdefense clusrer
:DSQ2. Bi~nd.1991 I W ~ mcluded S ;n Srcnd) 2 The DSQS contzins 1 3 : t e m s
intended to measure s ~ l i t i l c g(3 :[ens; or other closer! related rrrechanims
of d e i e ~ w Hr:. this \tuck. partiicipnts raeed t h e x agreement or drsapreemenr
for e x h of the DSQ2 items on a 5-pcinr Lrkeri-type scsie

Results and Discussion

[email protected],Vj/. The test-retest correlations fi;s rhe EB and its s ~ b -

scales (r; = 3 1 ) were ail highly sigcificant {(p< .0Ol ): SH, r = ,861 Self, r = .E3;
Family. r = .90: Others. r = .?5. These :es~i';s ;r?dicz.tethe SI arid its sobscaIes
have a high degree of temporal stability.

i.'~femaf eor?srstenc)/ The results of the xiiabmlity anaIjsis ( $ 3 = 79:

~ r a d u c e darr atphs cneff~cients f .92 fcrr the SL, .P8 for the Self fa~ro:. sf?
for the Famiky factor, and 85 for :he Others k t a r . These @\timakes of
re!,iabllrty are u n ~ f n r m l yhigh and cc:n\istenr with those found in Srtldy 1
In addit:on. rhe iukxcale rntercorrelations (presented In the lower half of
Table 3 ) were high:y significant <p < ,001 1 snd conszstent w r h the resnits
found in Study 1: Self and F a m ~ l y .r = .33: Self and Others. r
Family and Other<. r = .60
-
56;
 Gender differences. Based tin the sample of 78 participants ivhn indi-
cared their gender. !he difference in SI scores be:weer, mer, (M = 2.04)anci
women rM = '1.25) did not reach significance. r(76' = -I 2 2 . effecr size = -.Jli.
There was a nonsignificant trend in :he difference betwem n a i e iM = 7 56'1 a d
fcnzde (M= 2.98)scores on :he Self factor, :(76; = - 1 .X9.p < ,113. efiecr size =
-.41. There were no signitkarm: differences betumeen m d e [Jf = 1.5?: and fernale
::I4 = I&&) smres on the FarnjZy factor. ri'76i = - - 1 , 3 S . effect size = -.32. or
between male ( M = 1.9'3 and femaie !!?f = 1.92) scores on the Otbez factor:
::76>= ,311, effect. sIze = .R.The gender differences on the Se!C and Others
Downloaded by [Central Michigan University] at 17:14 04 January 2015

sabscaks faun2 in Study i were not repBii.a;ed in Srnciy 2.

Construct vaiidity, 08niy cclnvergerrt validity measures were ~ncluded

in Srcciy 2. and these resuits are presenied in Table 6 . Forty-tu.0 participants
conny.leted the first questionnaire packet used in :his stuciy. w h i c h c o n t a ~ n e d
the BSI, STB, and MCM-II measures. The SI and i:s scbscales were
significantly reiared to these m e a s e r e of the bnrderli!>e personslit\, dism-
der. !his replicating the finciings invo!ving [he BSI and STB measures
found ic Study 1.
Thirty-seven participants compbered the second ques~ionnairepacket.
which contained social ps>chologicai measares of s e l f - c o n c e ~ itabilit!
t
and se!F-esteem as weii as the DSQ2 measure of i r n s g e - d i s t c r i de-
fenses. -4s predicted. the SI and its subsclties had significant !~egati.;e
correlations with the RSS, RSE, ai:d TSBI measures. Therefore. :he SI
does appear to be associated w i t h l o w s e l f - c n ~ c e p t \ t n b i i i t > . anci Ion,
seif-esleenn, Howeve:, the S i and irs s u b x a I e s did ni?r correirtte signifi-
cantiy with the DSQZ measirre. Examining the ciegree of assacia:ion
between the SI and the 3 - i t r n spjitring indiimiduai defense scale of the
DSQZ yielded a significant correkatinn with :he Self fnct-or only., r = .ZR.
g < .G5.

TABLE 6
Gonvergen: Validlty Coefficients for Stuuy 2
Scaiv BSI STB IWCMI-II BSQ~ xss RSE TSBI

Sphttlng Index .68*** .60*"* .6Sa** - .% - 66*** - .68*** - !I**

Note n = 42 for BSI, STIZ, and h.1CMi-I:. n = 3" For DSQZ,RSS, RSE. and TSBl Scak
key. BS1 = Borderline Syndrome Endex; ST& = Schzotlpal-Borderrlne s d e . MCMI-I1 =
hG%ionCiinicd MuBt~axid Invento;).-11, Borderline Personality D~sorderScale, DSQ? =
image-dlsrorting defense style of :he Defense Style Questionnaue. RSS = Rosenberg Stabl!r:>
of Self scaie. RSE = Rosenberg Self-Esteem siaie: TSBI = Texas Socld Behmior Inventon.
* p c .05. **p < .Ot *"p < OOI
 This set of studies provides encouraging preliminary data for the SI: a new
clinical Instrument for assessing the splitring defense ~nechanism as de-
scribed by Keraberg (1975, 1976. 1984). The SI appears to be a reliable
measure, as d e m ~ n s t r a t e dby its high degree of icternai consistency. stability
over time, and meaningfuI and stable internal structure. The f i m l version of
the SE contains 24 items and three 8-item factors: "Splitting of the Self
fmage;" "Spliraing of Family Images," and "'Sppkitring of Others' Images."
The construct validity af the SI was also supported. The S1 had uniformly
Downloaded by [Central Michigan University] at 17:14 04 January 2015

high positive correlations with three measures of the borderline perscrnaiity

disorder and a measure of the narcissisiic pexrnality disorder. The ST had
reliable negatiw correlations with social psychoiogicai measures of self-
i n a g e stability and positive seif-esteem, demonstrating t h ~ the
t S I is associ-
ated with iderrtity diffusion and low self-esteem. The SI also had reliable
positive correEations w::h measures of depression and negative ai'fectiviiy
and a reliable negative correEation with positive zffecf viry. suggesting the SI
is associated with feelings of depression and anhedonia. Also consistent with
predictions were near-zero correlatioas with two measures of cognitive corn-
plexl:~. demonstrating that the SI does not simpik- measure undifferentiared
or rigid thinking.
Contrary to predictions. the ST had a snxaXI b i ~ tsignificant positive corre-
larion wirh rhe Dogmef sm Scale: a third measure of cognif ve complexity.
This result- is consistent with research suggesting h a t dogmatism is assocl-
cited wirh % O Wseif-esteem and negative v i e w of thc self and others (Ehrlich,
1978,;. The SI also faiied to correlate as predicted with the measures of
splitting and related defenses from the DSQ (Bond. 1991j. This finding is
consistent with the inability of Bond and his associates (Bond. 1990; Bond
et a]., 1989; Bond & Vaillant, i 980) to obtain validaring evidence for these
scales of the DSQ.
Xdditiond research is necessary before the SI may be confidently used as
a measure of the splitting defense mechanism. The SI was deveIoped and
vniidated on a popuIatlon of cojiege students. It r e x i i n s to he seen whether
the SI would be elevated in a clinical sample of patients believed ro rely on
the splitting defense mechanism. Future research should csmpare {he SI
scores of a sampje of borderline persenaiiry disorder patients 10 these of a
clinical s a ~ p l not
e predicted to use the defense mechanism of splitting. such
as a neurotic or obsessive-compulsive sample. as weII as to the SE scores of
a normal comparison group.
Second, all of the validity evidence in the set of studies presented here
used self-report measures. Campbe1.I and Fiske (1959) argaecl that measures
of unrelated traits n a y in fact be related due ro the sharing of measurement
features. Campbei! and Fiske reconmended using a "multitrait-nsu!tlnethod
matrix'' in which related and unrelated traits are crossed with different
merhods of measurement. For example, the SI could be compared to the
 Lerner and Lerner Scaje ( I x r ~ e S. r Lerner, E9S9: for assessing splitting on
the Rorschach. A theoreiicaiiy unrelated trait meastlred hy both the self-re-
port ond :he Rorschach methods should aiso he included. If w o d d thei? he
predicted that the sp1it::ng measures using different methods iiauld be more
highly correloted than the unrelated traits using :he s a x e method>. I n :his
manner. the effects of method variance could be c o n m l l e d in asaesiiing the
.i:aliditj. of the ST.
O x e jralidated with approp~iatecljnicaI poguiadons. the SI should ha\-e
important practical appiications in both sesexch and clinical settings. .A.
comn?onlq cited criticism of psychoanalytic theories is that they are dif5c~1lt
Downloaded by [Central Michigan University] at 17:14 04 January 2015

, .
ta examine emplricalk i:e.p.. Jaff'e: i99O'). The SI may provide zn object:\ e
measure for in~estigatingpsychomalyric oh-ject-relations theories relate2 to
the splittir:g defense mechanism. For exompie, Kernbe-p (T9?C) b e l i e ~ e i !
t!aa': the consistent inierpreration leading to the reductinn af splittiilg rnecha-
!:isms j s c x c i a l to the suceessfui treatment of borderii2e and narcissistic
perssmlity disoxkrs. He also stateci that splittir,g I s z ceiltrei mechanism in
the mere severe personality disorde-s &mi.be:g. 1975, !983'i. These nnd
other teneis related to spEktting could be ;ni;estigated x i n g the ST.
The SI ma! p a v e valuable as an easily administered screecing device I F
a clinicd setting. Many authors ha\e emphasized the difficulties :hat arise iii
the hospital treatment of patients with a heavy reliance n:1 spiirting (e.g.,
Gallop. 4'85; Kernberg. i984, Masterson. 197%).Spiitting n?ay also ha\.e an
adverse impact on individaai psychotherapy if rhe therapist does tior rtxng-
nize the patient's reliance on ':his defense rnechrnism and its abiijry to c r e x e
e x t r e m e cc~untertrunsferencereactions. In addi;lon. eieva:jons <,r! pnrtic~iar
s;l&scales crf the SI, such as the Family factor. could indicate problem areas
and serve as a m e f d guide to rreatrnent focus.
Finoily, the SI shot~ldbe a usefnl tool for diagnostic ptrrposes. The prs-
~orniraanceof spiitring and related mechonisrns signifes a se\:ere level of
psyohopathiogy with important consequences for appropriats therapeutic
stra!egies (see Akhtar 6 Byrme. 1983). As part of a psychodiagnostic batter):,
for example: the Sk couid provide important information regarding the diag-
nosis of a severe personali~yiiisorder.
From a theoreticai perspective. this set i>f r:ubies offers prelirnir:ary
s u p p c ~ t i i . eevidence for Kernberg's i l 9 7 5 , 1976: 1984: object-xiations
theory it: relation to the splitting d e k c s e mechanism cocst-act. The Inter-
ne! structure crf the SI, demonstrating a division of spiitting items be-
tkveen serf and object represen.:ations. is consistent with Kernberg's
ciescription of the splitting m e c h a ~ i s mof defense. -4: the s a n e rime. the
temporai stability of the SI, which measures unstahfe znd d i s t ~ r t e dper-
ceptions of self and others. supports Kernberg's (1975'; asser:ioz thar :he
borderline personality organization describes "patients tvho do have a
specrfic. srahle jitzlics added]. ~ a t h a i o g i c z Ipersonality organizarion:
their personajity organization is not a transitor? state Fiuciuzting beraesr:
Reurosis and psychosis" (F. 3).
 l a a d d ~ t ~ oKernberg
n, 1 1 9 7 3 emphaslzeci :he promnence of the y l l t t m g
mechacIsm rrr borderline and narcrssis:ic personality disorders, whzch is
corroborated b! the our resuits He rargaed that spirttlng 1s a central m e ~ h a -
nism EP, 3 %jade range sf severe psycbopathologq that 1\ cmsistent w ~ t hrhe
rehzb:e posjtjve correlations found between !he SI and :he nxxmx-es of
dcpressror: and neg3e1.iTeaffectrvmty Kerrrberg ( E 975 : alao behewd rejiznce
on the spl~ttrngdefense rnechanjsrn lead5 to rdentrty drffuseon and low
self-esteem The re!:able rregatlve. correlatrons Seeueen rhe SI and the mea-
\ures of self-image stabihty and pos:tIve \elf-esteem le:nfc~cethis w r l i o n .
In szmrnary. i h ~ csei of seudres offers encouraging ers,prtrcd evidence for
Downloaded by [Central Michigan University] at 17:14 04 January 2015

the rel~abilrrjaqd 1 ahdity of the SI However. these result., should be \ l e ~ , e c i

as grellrn~nary.Add:mnai research with psychopacholog~caigroups known
to engage in spl~ttingmust be conducted before the S1 can be conLrdentiy
used as a m e s u r e of the sp1:tarng defense mechanssrn

ACKNOWLEDGMENTS

Thr\ article js travd or? the doz:oral ciissertat;on by I c h n R. Gould 1Gor;id.

1993) for whizh Norman M Prentice and Ricardo G Amshe s e n e d as
cd-cham
Ap;.recraeion is. expressed to the remsining committee members. Randy L.
Dlehi, Robert L. Heimreich, and Robert X Young. for constructive s q g e s -
:Ions throughout the research. The assstance of Stephen Frnn and Gerald
Metaisky in the earfy phase of the srudy rs gra:efdI~ acknowledged.

REFERENCES

Adornor a. W., FrerrkeI-Brunswik. E.. tevmson, D. .!.. 8 Sanford. R. W . ii950). The au:hrpnti:r-
iun ~ r e r s o n d i ~Ncw
. 'dork: Harper 8 Brothers.
likhtar. 5.. d Byrnc, J. P. (i9t33c.The concep; of splitting znrd i s chnicrtl relevance. ."irner~m
Jcrurnrr: ($Psychiatry. !&IOi I.3-lOi 5
American Psychiatric Association. (i980)Diagnasric and .srar;s;;cai rnunual of trirnlu! disor-
ders (3rd ed.). woshi~gton.DC: Author.
American Psychratrlc P1sso~iati3n.(i994). Diqcastrc arid src;ri.rtiixl mnnurr! (!f men!d disor-
ders :4th ed.). U'rashngton, DC. -4uthor.
Andrcws. G , Poliock. C.., S. Siewari, G . (1989). The dererminstim of drfc-nse s:yk hy ques:run-
naire Archives :ifGenerai F.~)diiaf.ry, 46,455-450
Ashby, IH. U,. t e e , R. X.,B Duke, E H . i!94?, Sep!.embes). A r!~rcissi.tiicper:wni;lir). drsorder
MMPI srule. Paper presented at the mee:ing of the Amerrcan; Psychological Assocrotion, h'ew
York.
Beck. .4. T., B Steer. R A . (1987). Mtrr~uai,fa:? iizr f2er:seci Bfck Depressrirr: i r i v r n t o ~S S ~ E
Antonio. TX. Psychoiagjcal Corporation.
Bieri, J !1 9 5 9 Cngn~rlvecompiexity-simpiicity and preilictzve behavior 9olrrnd q! Abn:!mmd
ucd Sorkl Psychoiogy, 3?, 263-268
Bond, M. (1990). Are "hordrrline defenses" specific for borderline yersofiality d~sorders'
Journcii of'F'ersr~tm'i?.yDisorders, 4, 251-756
 bond. M. i i W i i. Mrrmccli,ii~c!he D p j r r i ~ rSryir Qucs:iorrmi;re UfipuSiichri rnanuscrrpt. Depart-
men: of F,.ychiatry. S:r Moriimer I3 Dsrrs-Jewsh General Fios~ita!. hlonfrrnl Canada.
Lion;. h l . Gardner. 5 7 . Christian. I.. k S!gal. ! i. ilPb3! Eniyir~csi.irud! of >elf-rstci
defense styies .Anh!;,ri (I! Gt.r;prcii i'.r~;.h~:i.~, $5, 733-738
liund. 54.. PC?:?. C.. Gauiier. h l . . Goiienberg. hl.. Oppenhcimer. i . k S~manri..i : i5*L!
Validating the seif-repor~cC defense sr)les. ji)!irririi I!? ?t r:.rbwiig. D:ro~-dt~r.. 3. !GI- i i 3.
bond,. 3.1. B Vaiilan:. I S. r 1066!. An ernpiricsi siuti) of the rsiarionhnp 5e:ween di-gnoslr and
defcnst. styie. .J~-r!i:vu: nf'Gencn:I ?.syi:~~rirv43 1 5 - - 2 8 R
Budner. S . i i962i Iniolerance o f arnh:guity as a peracinairty \,ariaaie. .:i::rr.r;::l irt P c r ~ ~ ~ n : i i r r .
j(j, 29-53,
Cr;rnpbell. D T., B F i s k . 3 ii' ( 1 3 3 i Convergent and discrmlinan! sairJa:j;vi k.4. the muit)-
Downloaded by [Central Michigan University] at 17:14 04 January 2015

tral! --nru1:imethod ma:r:x Ps?cti;dry;io! BlO!eii~i, 5 5 . 61-f O

C 1 m d g r , G.. 6 Brcks. P !8484) Schizorypy and hemisphere func:i:>r.-I Thrcr-tlcai con5:d-
erstion? and :he measuremen1 3f ~ i h i ~ ~ ! ! p j . PPTS:III~!:>
. c i r i i I9i&1.;c:'kd:1! 3i;:krec:e: 5.
S?i-SS$.
.
C o n k 11. K.. Plcrchik. R . K x a s u . T. E & J e r e t l . 1 i i4W! X self-report kmrdsriinc. scale.
g i r l cise. ;6h'
m:.; .rf~::;~~,!
D i s c r m i n z i i w vat!dity and preiinl:nar? norms. Joiirtiiil (1; ~~:er>,r?!~s
318 4 3 4
Cronbach. L.J . [ i?S! I Coeffictent aiphs and the internal strc:rure of teirs P.r\c/~iwrc.r.vhj5,
"'77.34.
Ehrlich. El. 1. i197S!. Ihgmarisrr.. In H L~ndo:: k J E Esner. 3:. (ECS.). D ( m r u ! : i ? : so?
nrvc~rrc;;~& !pp. 365-3%). New York. W l e y .
Gaiiop. R . :'i?h5>. The patient is spi~tting:Everyone knows and nothing :hanger Jr,llrnii; rli
F:!t-hrrsoc;:;l .?iu:s:rig oi;d M e c i s ! HeaLri; Scr\.rcrs. 23. 6- 10.
Grrsun, AC J ( !??A). Splittmg. The dereiopmrn! of a rneasurs $:~urti,i, t l f C'i!uiL.;lt Pr~c.!rv(r~i.:, ..
&G :.F7-;6:.
G!avacch:ni. P L. !iOB.A.), Ciiarartr: d:sordrrs a n d udqr:!i.e me:iu:rl,~ii.r Norrb\nis, NJ
ksnson
Class. G \' t 1q371 1n:rgra:tng findings. The rrretn-andysi: o i re\e:ir:i;. K c i . , o r :,i K p i e ; i . r . - / ; :rr
Edfica:;clr!. 5. 351-179.
Giassrnaz. M. 5 I i9S6). Spl!t:ing. Fiir:her con:rihuiions ti-, :he d r r t i : i ~ : ~ ~ e ni:f'r a meaqure
Jourm:! q? C!;t;t:.c:! P.?c h o l o p $2. 895 -904
Giluid, :. R . i 19931. The development and validation of 2 bi.;'ie rnrasmn: !he defense n u h a -
nisn: o f spiirring <Docram! dissertation. L'ni\,ers!ty of Texas at 4usrir:. I L?'.?,!, D;s.FEI.:c:~~.II:
Ahstrcrc !s Intemurionni. 54. 266B-2267R
Gundersoc, J E. ! iSiS7). Interfaces betwecr? p ~ y c h o a n a i y sand empirical studies of bo:dcrline
personaliiy. !n J 5 Grotstein. hf F. Sofomon: 6 !. 4 . t a n g tEds 1. Tile b;,?;ieri;ric plximr.
Erner$ctq t.~ncq?:.r in 1!:~2gn2.~i.7, ; ) ~ ~ t ~ h c ~ i ; ~ ~ ! ~and~ r t ; rwc~rm61:i
:c.?, (1.0; i pp ??-!G:.
Hillsdale. Anaiytic Press.
Eunderson, :. G.. & Singer, hl T, !.!V5:. Defin?n,o boriieiflne pntirnrs. Ar: a ~ e r v i r vAwericcr;
i:~ic~:iii<:! P.s?c.!:ruin., i3Z. I -.1:'.
Hathnwa!.. 5. R.. 6:M c K i n l e . :. C. r IS.;?). Ti:c .\l:n?lesi~ia,Vlii:iph(i:: .f'rr.rorm(:!~ iristniw:,
Minnenpohs: L1,'niversit? of Minnesate Press
Heimreich. K., Srspp. I.. lic Ervin. C. t l974j. T h r Texas Socrcil Rehs\.i;,r !nwncory :.TSB:!. .Ar:
o b j e c t ~ v cmeasure oh self-esteen? or socval competence !ouri:i:i S!~p:,ier?:c?i:Absrriic: Sev:c-r
Ccri~ii:!r(t! Seieciea' Doc.t:~tma,cr,: P.cyclioic;p. 4 i 2 . l ~N o 6&1!. 79.
;;~ffe. L S , (i9W;t. The rmpiricd foundntions of psycbomalytic 3ppr33:39!: LO psyc.hr.jogicd
resting. .?oi:rr~c:cf'Perscwai;i?. Assessmri:~.5-7. 746--'55
Kzrnherg. 0 F. (1975:. Border!ice ;.;,ndi:i:*r?s c~r.c'pihoiogrcu!n:irc:ssi;rfi. Nru York .4:(?nw::
Kenberg. 0 F. :197h'.. Oi,jei-: reia!;cw tiret!.q onr1ciii:iri;l p.:~:.h;~iincii~iiiS e i s York: Aronsor..
Kernberg. C ) . F (1384). .%.ert prrsor:ni:? disc!sders: ,~r:~~c~!io;i:ercpc.i.i~( r:.riiii.g:cx. K e n
Waver.. CT. Y a k Uriiverslr!. Presb.
 Kohut. H. (I971!. The atznl;\.sis of the self. New York. Intrrnationa! Unlvcrsities Press
Lerner. F. M.. & Lerner, 11 D. (1980). Rorschach assessment of primitive defcnses in borderline
personality structure Is, J. Kwawer, K.Lerner, P. Lerner. & ,4. Sugarman (Eds.). BnrderElr;~
pi~enomerra and the Rnrschnch r p s r {pp. 3 7 - 2 7 4 ) . New \hrk: international Universrties
Press.
Mnrmar, C. It.. X. Horowitz, M. 3 . <1986i. Phenomenoiogicnl anaiysis of splitting. P.~;~chalkrr-
up): 2.1. 71-29.
Masterson. J . F. :1975). The spirtting defense mechanisn? of the b o r d e r h e adolescent: Devrl-
opnental an& cilnicai aspects. Ir; J. E. Mack { E d . ) , Bnrderiinc sia1e.r Ir: p.sjch!arry !pp
93-.;0l j. New York. Crune K;. Stratton.
Downloaded by [Central Michigan University] at 17:14 04 January 2015

John R. Gouid
C/O Norman: M. Prentice

Received September 9,1394

Revised J u n e 1'1, 1 9 9