O iginal Pape
Desensi izing Anxie y Th ough Impe cep ible Change:Feasibili y
S udy on a Pa adigm o Single-Session Exposu e The apy o
Fea o Public Speaking
Domna Banakou1, BSc, MSc, PhD; Tania Johns on2, BSc, MSc, PhD; Alejand o Beacco1, BSc, MSc, PhD; Gizem
Senel1, BSc, MSc, PhD; Mel Sla e 1, BSc, MSc, MA, DSc
1E en Lab, Depa men o Clinical Psychology and Psychobiology, Ins i u e o Neu osciences, Uni e si y o Ba celona, Ba celona, Spain
2Sys ems Neu oscience, Augus Pi i Sunye Biomedical Resea ch Ins i u e, Ba celona, Spain
Co esponding Au ho :
Mel Sla e , BSc, MSc, MA, DSc
E en Lab
Depa men o Clinical Psychology and Psychobiology
Ins i u e o Neu osciences, Uni e si y o Ba celona
Campus de Munde Edi ici CAVE
Passeig de la Vall d'Heb on 171
Ba celona, 08035
Spain
Phone: 34 93 403 ex 9618
Email: melsla e @ub.edu
Abs ac
Backg ound: Exposu e he apy (ET) o anxie y diso de s in ol es in oducing he pa icipan o an anxie y-p o oking si ua ion
o e se e al ea men sessions. Each ime, he pa icipan is exposed o a highe anxie y-p o oking s imulus; o example, in he
case o ea o heigh s, he pa icipan would successi ely expe ience being a a g ea e heigh . ET is e ec i e, and i s coun e pa ,
i ual eali y (VR) exposu e he apy (VRET), whe e VR subs i u es eal-wo ld exposu e, is equally so. Howe e , ET is
ime-consuming, equi ing se e al sessions.
Objec i e: This s udy aimed o compa e he esul s o single-session exposu e wi h hose o adi ional VRET wi h ega d o
educing public speaking anxie y.
Me hods: We in oduced a pa adigm conce ned wi h public speaking anxie y whe e he VR exposu e occu ed in a single
session while he pa icipan in e ac ed wi h a i ual he apis . O e ime, he he apis ans o med in o an en i e audience wi h
almos impe cep ible changes. We ca ied ou a easibili y s udy wi h 45 pa icipan s, compa ing 3 condi ions: single-session
exposu e (n=16, 36%); con en ional mul iple-session exposu e (n=14, 31%), whe e he same con en was deli e ed in successi e
segmen s o e 5 sessions; and a con ol g oup (n=15, 33%), who in e ac ed wi h a single i ual cha ac e o alk abou e e yday
ma e s. A week la e , he pa icipan s we e equi ed o speak on a s age in on o a la ge audience in VR.
Resul s: Ac oss mos o he se ies o con en ional public speaking anxie y measu es, he single-session exposu e was a leas
as e ec i e in educing anxie y as he mul iple-session exposu e, and bo h hese condi ions we e be e han he con ol condi ion.
The 12-i em Pe sonal Repo o Con idence as a Speake was used o measu e public speaking anxie y le els, whe e highe alues
indica ed mo e anxie y. Using a Bayesian model, he pos e io p obabili ies o imp o emen compa ed o a high baseline we e
a leas 1.7 imes g ea e o single- and mul iple-session exposu es compa ed o he con ol g oup. The S a e Pe cei ed Index o
Compe ence was used as a measu e o an icipa o y anxie y o speaking on a s age in on o a la ge audience, whe e lowe
alues indica ed highe anxie y. The p obabili ies o imp o emen we e jus o e 4 imes g ea e o single- and mul iple-session
exposu es compa ed o he con ol g oup o a low baseline and 489 (single) and 53 (mul iple) imes g ea e o a middle baseline.
Conclusions: O e all, he esul s o his easibili y s udy show ha o mode a e public speaking anxie y, he pa adigm o
g adual change in a single session is wo h ollowing up wi h u he s udies wi h mo e se e e le els o anxie y and a la ge
sample size, i s wi h a andomized con olled ial wi h nonpa ien s and subsequen ly, i he ou comes ollow hose ha we ha e
ound, wi h a ull clinical ial wi h pa ien s.
JMIR Fo m Res 2024 | ol. 8 | e52212 | p. 1h ps:// o ma i e.jmi .o g/2024/1/e52212 (page numbe no o ci a ion pu poses)
Banakou e alJMIR FORMATIVE RESEARCH
XSL
•
FO
Rende X
(JMIR Fo m Res 2024;8:e52212) doi: 10.2196/52212
KEYWORDS
exposu e he apy; i ual eali y; g adual exposu e; ea o public speaking; anxie y; change blindness; public speaking;
desensi iza ion; anxie y diso de ; easibili y s udy
In oduc ion
Exposu e he apy (ET) o anxie y diso de s (and o he men al
heal h condi ions) in ol es he sys ema ic desensi iza ion o e
ime o pa ien s o ea ed s imuli o si ua ions. Fo example,
pa ien s wi h ea o public speaking migh be slowly in oduced
o si ua ions o public speaking: in he i s session, pe haps o
a pho og aph o an audience, in he nex o a ideo, and hen
g adually building owa d speaking o a li e audience. This has
p o en o e se e al decades o be an e ec i e he apy [1,2].
Vi ual eali y (VR) exposu e he apy (VRET) uses he same
echnique, bu VR p o ides he s imuli, elying on he ac ha
people end o espond ealis ically o e en s in VR [3] and
he eby would exhibi anxie y in esponse o he ea ed
si ua ions. Fo example, in he case o ea o public speaking,
a p e equisi e o he success o VRET is ha people should
exhibi simila anxie y o a i ual audience as hey would o a
eal audience, which has been shown o be he case [4,5]. This
app oach has been s udied and used o e he pas 3 decades in
he ea men o a a ie y o phobias and men al heal h diso de s,
including bu no limi ed o social anxie y diso de s,
claus ophobia [6], panic a acks, pos auma ic s ess diso de ,
ea ing diso de s [7], and gene alized anxie y diso de s [8,9].
Ea ly examples can be ound in Ro hbaum e al [10-12] and in
Di ede and Ho man [13], wi h a e iew o his ea ly wo k in
K ijn e al [14] and a me a-analysis in Powe s and Emmelkamp
[15], which showed a la ge e ec size o VRET compa ed o
con ol condi ions and no disad an age compa ed o in i o
ea men s. This was seen again in a andomized con olled ial
(RCT) ha speci ically conside ed social anxie y diso de s [16].
Mo eo e , VRET has p ac ical and logis ic ad an ages since
he en i e ea men can ake place in he o ice o he clinician,
a he han a anging o ou side isi s o he pa ien s o be
exposed o eal-li e e en s (eg, isi ing a high loo o a building
in he case o ea o heigh s o being exposed o a li e audience
in he case o public speaking anxie y). Mo eo e , he e is
e idence sugges ing ha VRET may be p e e ed o in i o
he apy [17] by pa ien s wi h anxie y diso de s. Recen
me a-analyses ha e con inued o show ha VRET leads o
he apeu ic ou comes ha a e a leas as success ul as in i o
ea men s [18,19], including cases o se e e anxie y, obsessi e
compulsi e, and pos auma ic s ess diso de s [20] and
speci ically in ela ion o public speaking anxie y [21].
VRET has been ex ensi ely esea ched in he ea men o public
speaking anxie y [16,22-25], whe e indi iduals ini ially
expe ience anxie y, leading o beha io al, cogni i e, and
physiological symp oms, when deli e ing o an icipa ing he
deli e y o a speech in on o an audience. ET o public
speaking anxie y in ol es g adual exposu e o indi iduals o e
a ious sessions o si ua ions o public speaking, whe e a each
successi e session, he si ua ion becomes close o speaking o
a li e audience. VRET ollows he same idea bu wi h i ual
audiences. VR, in pa icula , o e s g ea e lexibili y a low
cos in he sense ha aspec s o he s imuli can easily be
changed: audiences can be o di e en ypes and be eac i e
[5], and pa ame e s such as he size o he audience can be
changed [16,26,27], including hei appea ance and beha io
[5,28], as well as he i ual loca ion and con ex o he speech
(eg, class ooms, lec u e ooms, con e ence ooms) [24] and he
ep esen a ion o he sel [29,30].
ET, whe he in i o o in VR, can be accompanied by cogni i e
he apy, whe e he clinician a emp s o help he pa ien e ame
hei anxious hough s abou he si ua ion. Fo example, F eeman
e al [31] used a simple o m o cogni i e beha io al ET wi h
VR, whe e o e 5 sessions, pa ien s wi h ea o heigh s lea ned
o o e come hei ea . Thei ask was o mo e o a ious le els
o a building accompanied by a i ual he apis and a each
le el o pe o m some asks on a balcony o e looking an a ium
below. Ul ima ely, hey made i o he op loo , and hei le el
o anxie y was signi ican ly educed compa ed o be o e he
exposu es and o he con ol condi ion.
A me a-analysis by Lim e al [32] ound an a e age o 6 VR
sessions, each las ing a ound 37 minu es, among e ec i e VRET
in e en ions. Ano he me a-s udy by Hinojo-Lucena e al [33]
ound ha up o 12 sessions a e necessa y, sp ead o e a week
o e ec i e ea men . This is simila o he inding o Ree es
e al [21], who epo ed be ween 5 and 12 sessions in he pape s
s udied o be e ec i e, while Chesham e al [18] ound up o
14 sessions o be e ec i e. Al hough no in he con ex o an
ET in e en ion, Boe je and an Ginkel [34] a gued ha he
op imum numbe o VR sessions o educing public speaking
anxie y canno be p esc ibed, since he cha ac e is ics o VR
ea men s a y among he e iewed s udies. Following om
his, Lim e al [32] a gued ha he numbe and leng h o sessions
should be conside ed as a unc ion o se e i y o he pa ien ’s
condi ion.
He e, we in oduced a new pa adigm based on he idea o VRET
bu wi h a single VR session du ing which pa icipan s ad ance
owa d a si ua ion ha would cause g ea e anxie y, wi h changes
occu ing almos impe cep ibly. Pa icipan s wi h ea o public
speaking a e engaged in con e sa ion by a i ual counselo ,
who explains he issues behind ea o public speaking and
encou ages he pa icipan s o speak abou hei own public
speaking p oblems and gi es hem a ious exe cises o do. The
counselo is ep esen ed by a i ual human cha ac e , who
s ands in on o he pe son, acing hem, and alks and ges u es
in a na u al way. A e a while, a copy o he counselo eme ges
om behind hei i ual body, so he e a e 2 iden ical ins ances
o he counselo , al hough only he o iginal one con inues he
dialogue, while he o he one mo es o o he side and con inues
lis ening. As ime p og esses, his di ision p ocess con inues,
whe e he addi ional copies eme ge. In addi ion, a e a while,
he new copies g adually ans o m in o di e en i ual human
cha ac e s. This p ocess con inues mainly in pe iphe al ision,
JMIR Fo m Res 2024 | ol. 8 | e52212 | p. 2h ps:// o ma i e.jmi .o g/2024/1/e52212 (page numbe no o ci a ion pu poses)
Banakou e alJMIR FORMATIVE RESEARCH
XSL
•
FO
Rende X
and he changes a e impe cep ible. O e ime, he s anding
i ual cha ac e s adop a sea ed posi ion. By he end o he
session, he pa icipan s ind hemsel es speaking in on o
an en i e sea ed audience.
The idea behind why his app oach may be e ec i e is o
conside a which poin he anxie y p o oked by public speaking
would become ac i e. Fi s , he pa icipan speaks o only 1
( i ual) pe son. Then, he numbe o people becomes 2, bu i
is he same pe son— he counselo . Howe e , 2 is no an
audience, no is 3. I 3 people a e no an audience, a e 4? I 4
people a e all igh , does 1 mo e ma e ? Gene ally, i he
pa icipan is com o able speaking o n people, hen
impe cep ibly making he audience n+1 should make no
di e ence ( o n>1). In addi ion, he changes ake place slowly
(excep o he i s ), so he e is no ob ious momen when he
pa icipan is no speaking o an audience and hen is speaking
o an audience. The hypo hesis is ha people can lea n h ough
his p ocess ha jus as i is possible o speak o 1 pe son wi hou
anxie y, he g adual ansi ion o an audience should no gene a e
anxie y. The e is no poin o discon inui y, whe e a one momen
he pa icipan is suddenly speaking o an audience bu a momen
ago was no . The hypo hesis is ha his lea ning will ca y o e
o subsequen speeches in on o e en la ge audiences.
We ca ied ou an expe imen o explo e he u ili y o his new
pa adigm. The pa icipan s we e no pa ien s bu people who
epo some le el o ea o public speaking. Hence, his was
no a clinical ial bu a easibili y s udy o assess he e icacy
o he pa adigm. Hence, he speci ic aim o his pape was o
in es iga e whe he he g adual change pa adigm migh p oduce
a educ ion in he ea o public speaking ha is a leas as good
as he adi ional mul iple-session VRET app oach.
Me hods
E hical Conside a ions
The expe imen was app o ed by he Comissió Bioè ica o he
Uni e si y o Ba celona (IRB00003099), and pa icipan s
p o ided w i en in o med consen . All me hods we e pe o med
in acco dance wi h ele an guidelines and egula ions.
Expe imen al Design
We conduc ed he expe imen wi h 45 pa icipan s using a
be ween-g oups design, wi h a single ac o (exposu e) wi h 3
le els: single-session exposu e (n=16, 36%), mul iple-session
exposu e (n=14, 31%), and con ol g oup (n=15, 33%). In he
single-session exposu e, pa icipan s expe ienced he scena io
as desc ibed in he In oduc ion sec ion—whe e 1 counselo
e en ually mo phed in o an en i e audience. The
mul iple-session exposu e g oup expe ienced adi ional VRET,
wi h 5 sessions, each wi h an inc easing audience size o e
app oxima ely 3 weeks, wi h 2 sessions pe week (Figu e 1).
The i ual en i onmen , i ual counselo , and con e sa ion
we e iden ical in hese 2 g oups. Addi ionally, he e was a
con ol g oup, whe e pa icipan s con e sed wi h a
gende -ma ched i ual human who asked gene al ques ions o
he pa icipan s, such as hei name, job, skills, and in e es s
(hobbies), and also alked abou hemsel es (Figu e 2A,B). The
pu pose o he con ol g oup was o check ha simple exposu e
o any VR whe e he pa icipan was equi ed o alk would also
be su icien o educe anxie y.
A he end o he expe imen (single session, mul iple sessions
a e he end o he i h session, con ol g oup), pa icipan s
we e asked o e u n 1 week la e , when hey would be on a
s age in on o a la ge audience in a hea e in VR and would
be equi ed o in oduce a pe o mance by he band Di e S ai s
(Figu e 2C,D). This inal exposu e was o es ing. A his o ic
(1980s) band was chosen so ha pa icipan s would need o
conduc some esea ch o ind ou mo e abou i . Mo eo e ,
since he p esen a ion in on o a la ge audience was o be
se e al days la e , we could assess an icipa o y anxie y.
The expe imen al scena ios a e illus a ed in Mul imedia
Appendix 1, which shows a mo ie o he a ious condi ions.
Figu e 1. The scena io. (A) A male pa icipan embodied in a male i ual body wi h a i ual mi o o his le . The inse shows he pe son wi h he
HMD and con olle s. (B) Single-session exposu e: he i ual counselo alks o he pa icipan , while new copies o him di ide and g adually ans o m
in o di e en i ual human cha ac e s. (C) Mul iple-session exposu e (session 2): he pa icipan is exposed o 3 i ual cha ac e s o ming he audience.
(D) The ull audience in bo h single- and mul iple-session condi ions. HMD: head-moun ed display.
JMIR Fo m Res 2024 | ol. 8 | e52212 | p. 3h ps:// o ma i e.jmi .o g/2024/1/e52212 (page numbe no o ci a ion pu poses)
Banakou e alJMIR FORMATIVE RESEARCH
XSL
•
FO
Rende X
Figu e 2. Con ol condi ion and conce scena io. (A and B) Vi ual humans o he con ol condi ion. (C) Vi ual audience seen om he s age om
he iewpoin o he pa icipan in on o he mic ophone p io o in oducing he band Di e S ai s. (D) View om he audience once he conce
s a ed.
Rec ui men
Pa icipan s we e ec ui ed ia ad e isemen s a ound he
campus o he Uni e si y o Ba celona and h ough social media.
Pa icipan s had a mode a e le el o public speaking anxie y.
Du ing ec ui men , po en ial candida es unde wen a sc eening
p ocess using he 12-i em Pe sonal Repo o Con idence as a
Speake (PRCS-12) [35], a s anda d ins umen o assessing
public speaking anxie y whe e g ea e alues indica e mo e
anxie y, and he Liebowi z Social Anxie y Scale (LSAS) [36]
o iden i y hose eligible o pa icipa e (see Mul imedia
Appendix 2, which also includes de ails o inclusion and
exclusion c i e ia).
Ma e ials
We used a wide ield-o - iew s e eo head- acked,
head-moun ed display (HMD), h ough which he pa icipan s
we e embodied in a li e-sized (gende ma ched) i ual body
om a i s -pe son pe spec i e, spa ially coinciden wi h, and
isually subs i u ing hei eal body. Th ough uppe body
acking, he i ual body mo ed synch onously wi h hei own
body mo emen s. They saw a e lec ion o he i ual body om
he neck down in a i ual mi o and when di ec ly looking
owa d hemsel es. Full de ails o ha dwa e and p og amming
implemen a ion a e gi en in Mul imedia Appendix 3, which
also con ains he sc ip s o he dialogues be ween i ual
counselo s and pa icipan s.
Response Va iables
The e we e 2 ypes o esponse a iables. Fi s , a e he VR
exposu es, a ques ionnai e (“VR Ques ionnai e”) on he le el
o body owne ship and agency o e he i ual body and gene al
esponses o he i ual audience was adminis e ed. The
ques ions a e shown in Mul imedia Appendix 4; open-ended
ques ions ha we e used o sen imen analysis we e also
included. Second, we included he esponse a iables di ec ly
ela ed o public speaking anxie y, as shown in Table 1. The
“p e” p e ix e e s o adminis a ion o he ques ionnai e p io
o VR exposu e, “pos ” means a e he VR exposu e and jus
be o e he VR conce exposu e, and “a e ” e e s o a e he
conce . Fu he de ails o hese a e gi en in Mul imedia
Appendix 5 [37-40].
JMIR Fo m Res 2024 | ol. 8 | e52212 | p. 4h ps:// o ma i e.jmi .o g/2024/1/e52212 (page numbe no o ci a ion pu poses)
Banakou e alJMIR FORMATIVE RESEARCH
XSL
•
FO
Rende X
Table 1. Response a iables.
S udyIn e p e a ionVa iable
[37]
p eIATa, pos IAT •IAT o ea o nega i e e alua ion
•Range: –2 o 2
•Nega i e sco es: au oma ic p e e ence o sel / ejec ed, o he /liked
•Posi i e sco es: au oma ic p e e ence o o he / ejec ed, sel /liked
[38]
p ePRCA24b, pos -
PRCA24
•Communica ion app ehension: le el o anxie y igge ed by eal o an icipa ed communica ion
•Sco es: 24-120 (highe sco es indica ing mo e anxie y)
[39]
p eSPICc, pos SPIC,
a e SPIC
•SPIC
•Sco es: 15-105 (highe sco es indica ing eeling be e abou sel and conside ed as mo e compe en
pe o mance)
[40,41]
pos STAId•Sho - o m STAI adminis e ed be o e he conce speech
•Sco es: 8-32 (highe sco es indica ing mo e com o )
aIAT: Implici Associa ion Tes .
bPRCA: Pe sonal Repo o Communica ion App ehension.
cSPIC: S a e Pe cei ed Index o Compe ence.
dSTAI: S a e-T ai Anxie y In en o y.
P ocedu es
Du ing he i s isi o pa icipan s o he VR labo a o y, hey
we e gi en an in o ma ion shee o ead, and a e hey ag eed
o con inue wi h he expe imen , hey we e gi en a consen o m
o sign. Nex , hey comple ed a se ies o ques ionnai es assessing
hei ea o public speaking and nega i e e alua ion
(p ePRCA24, p eSPIC, p eIAT, and p eSTAI, whe e PRCA
e e s o Pe sonal Repo o Communica ion App ehension,
SPIC e e s o S a e Pe cei ed Index o Compe ence, IAT e e s
o Implici Associa ion Tes , and STAI e e s o S a e-T ai
Anxie y In en o y) ha also se ed as baseline measu es, as
well as o he demog aphic da a (see Mul imedia Appendix 6
o de ails). Be o e he expe imen s a ed, pa icipan s we e
i ed wi h a head-moun ed display (HMD) and body- acking
equipmen . The iew seen h ough he HMD was calib a ed o
each pe son. Pa icipan s in he mul iple-session exposu e g oup
e u ned app oxima ely 2 days la e o he second session and
so on un il hey comple ed all 5 ea men sessions. A e he
las VR exposu e, pa icipan s comple ed a pos -VR expe ience
ques ionnai e. One week a e pa icipan s’single (single-session
o con ol exposu e g oup) o inal exposu e (mul iple-session
exposu e g oup), hey e u ned o a ollow-up session, whe e
hey had o gi e he Di e S ai s welcome speech. Be o e
deli e ing he speech, hey comple ed he ques ionnai es
assessing hei ea o public speaking and nega i e e alua ion
again (pos IAT, pos PRCA24, pos SPIC, pos STAI). When hey
inished and came ou o he i ual en i onmen , hey
comple ed a pos -VR ques ionnai e and a ques ionnai e on he
pe cei ed index o compe ence (a e SPIC) ela ed o he speech
hey had jus deli e ed. Nex , hey we e deb ie ed abou he
pu pose o he s udy and compensa ed, and hey le he
labo a o y.
S a is ical Analysis
We did no make s a is ical in e ences abou he ques ionnai e
a iables ela ed o he VR expe ience (eg, body owne ship; see
Mul imedia Appendix 4) bu only wished o check ha he
esul s con o med wi h ea lie s udies. Hence, we only
conside ed hese esponse a iables a he desc ip i e le el.
Fo a iables speci ically ela ed o anxie y, we used a Bayesian
s a is ical model de ailed la e , equi alen o 1-way ANOVA,
o all bu 1 esponse a iable and a logis ic model o pos IAT.
This esul ed in pos e io dis ibu ions o each o he pa ame e s
o he model, om which we could compu e any p obabili ies
o in e es . The model included all esponse a iables
simul aneously, so he e was no issue wi h mul iple compa isons
ha would esul in p oblems in he in e p e a ion o signi icance
le els in classical null hypo hesis es ing.
Fo each esponse a iable (pos IAT, pos PRCA24, pos SPIC,
a e SPIC), he linea p edic o , which ela ed he independen
a iables o he mean o he esponse a iables, was o he o m
Condi ion + Co a ia e + (Co a ia e × Condi ion) [+ amilia i y]
(1)
whe e Condi ion e e s o he main e ec s o con ol,
mul iple-session, and single-session condi ions; Co a ia e e e s
o he p e a iables (Table 1); Co a ia e × Condi ion is he
in e ac ion e ec ; and he co a ia e amilia i y e e s o he
sco es on a ques ion abou he amilia i y o pa icipan s wi h
he band Di e S ai s. Since he SPIC a iables and pos STAI
ela ed o p epa a ion o gi ing a alk abou Di e S ai s, we
also included amilia i y o hese esponse a iables as a
co a ia e based on he esul s o Figu e S4 in Mul imedia
Appendix 7, which shows a lowe le el o amilia i y o he
mul iple-session condi ion compa ed wi h he o he condi ions.
The e was no p e co a ia e o pos STAI.
The linea p edic o o each esponse a iable was he e o e
o he o m:
(2)
JMIR Fo m Res 2024 | ol. 8 | e52212 | p. 5h ps:// o ma i e.jmi .o g/2024/1/e52212 (page numbe no o ci a ion pu poses)
Banakou e alJMIR FORMATIVE RESEARCH
XSL
•
FO
Rende X

This is a s anda d ANOVA model whe e:
•μ is he g and mean;
•
•αjis he main e ec o condi ion;
•γj is he in e ac ion be ween he condi ion and he co a ia e,
whe e j=1 (con ol), 2 (mul iple sessions), o 3 (single
session), wi h α1=γ1=0 so ha mul iple-session and
single-session condi ions we e compa ed agains he con ol
condi ion;
•Ciis he p e a iable as a co a ia e (eg, p eIAT); and
•Fiis amilia i y (used o pos SPIC, a e SPIC, and
pos STAI).
Since in he case o pos STAI, he e was no p e co a ia e, he
co esponding model was educed o he main e ec oge he
wi h Fi.
In he case o all esponse a iables, excep o pos IAT, we
used an ANOVA model wi h a co a ia e, so he mean (μi) o
he esponse a iable (yi) was se equal o he linea p edic o :
μi=ηi
(3)
In s anda d ANOVA, he likelihood (ie, he dis ibu ion o he
esponse a iable condi ional on he pa ame e s) is equi ed o
ha e a no mal dis ibu ion. He e, we can be mo e lexible and
le he likelihood ollow a S uden dis ibu ion. This has he
ad an age ha i has a wide dispe sion han he no mal, hus
allowing o po en ial ou lie s. In addi ion, o high enough
deg ees o eedom, i app oxima es he no mal, and o deg ees
o eedom o abou >30, i is indis inguishable. Hence, he
likelihood is:
yi ~ S uden _ (υ, μi, σ)
(4)
whe e υ>1 is he deg ees o eedom pa ame e s, μi is he mean,
and σ is he scale pa ame e . Smalle alues o υ and la ge
alues o σ co espond o g ea e dispe sion wi h espec o he
symme ic dis ibu ions a ound μi. Fo la ge υ alues (>30
app oxima ely), he dis ibu ion is equal o he no mal
dis ibu ion, wi h mean μi and SD σ.
In he case o pos IAT, his model did no p oduce a good i o
he da a. This is because IAT is bounded by he alues by
cons uc ion. Hence, we no malized he pos IAT and p eIAT
alues by ans o ming hem o a scale. Nex , we used he
Be a dis ibu ion o he likelihood: Be a was chosen because
he p obabili y densi y is bound o , and he dis ibu ion can
ake on many di e en shapes (symme ic abou 0.5, J shaped,
in e se J shaped, uni o m, e c) and he e o e can adap o he
da a. Hence, in he case o IAT, he model is:
yi ~ Be a(ϕμi, ϕ[1 – μi])
(5)
whe e ϕ>0 is a scaling pa ame e , and he mean o he
dis ibu ion is μi. To ensu e ha , we used he logi link
unc ion be ween he mean and he linea p edic o , wi h he
in e se:
(6)
This is a s anda d logis ic model.
We used weakly in o ma i e p io dis ibu ions (ie, p ope
p obabili y dis ibu ions wi h wide a iance [42,43]). All βj ~
no mal(mean 0, SD 100); hence, he p io 95% c edible in e als
(CIs) we e –200 o 200. All σ, υ, ϕ ~ Gamma(2,0.1); hence,
he 95% p io CIs we e 2.4-55.7.
All esponse a iables we e in he same o e all model, so he
pa ame e s we e, o example, deno ed by μp c, αp c, βp c, γp c,
υp c, and σp c in he case o pos PRCA24, and he o he s a e
shown in Table S4 in Mul imedia Appendix 8.
The analysis was ca ied ou using he p obabilis ic
p og amming language S an [44] wi h he s an in e ace in R
(R Founda ion o S a is ical Compu ing) [45]. In o al, 3000
i e a ions we e used wi h 4 chains, and all simula ions con e ged
success ully, wi h all Rha =1, indica ing ha he esul s o he
chains we e p ope ly mixed.
Resul s
Pa icipan S a is ics
The mean age o he 45 pa icipan s was 26.1 (SD 7.36) yea s,
27 (60%) o hem iden i ied as emale, 14 (31%) as male, and
he emaining 4 (9%) as “o he ” o “p e e ed no o say.” The
de ailed demog aphics a e p o ided in Mul imedia Appendix
6.
Desc ip i e S a is ics o VR Ques ionnai e Resul s
The median le els o body owne ship and agency we e high
and in line wi h he esul s o p e ious expe imen al s udies (eg,
[46]), al hough body owne ship and agency we e o pe iphe al
in e es in his s udy, and he pa icipan s only saw hei i ual
body in a mi o o a sho pe iod p io o he discussion wi h
he i ual human and p io o in oducing he band Di e S ai s
in he second phase. Posi i e esponses o he audience we e
middle o high in he single- and mul iple-session condi ions
bu somewha lowe in he con ol condi ion. Full de ails and
analysis o hese esponses a e a ailable in Mul imedia
Appendix 7.
Desc ip i e S a is ics o Anxie y-Rela ed Resul s
Table 2 shows he means and SEs o he esponse a iables ha
we e di ec ly ela ed o public speaking om Table 1. I also
shows he e ec sizes (Cohen d) o he di e ence be ween pos
and p e measu es and he e ec sizes o he pos -p e di e ences
compa ing be ween condi ions. The pos -p e e ec sizes o he
con ol g oup we e all small, excep o a e SPIC, which had
a medium e ec size. Compa ing he single- and
JMIR Fo m Res 2024 | ol. 8 | e52212 | p. 6h ps:// o ma i e.jmi .o g/2024/1/e52212 (page numbe no o ci a ion pu poses)
Banakou e alJMIR FORMATIVE RESEARCH
XSL
•
FO
Rende X
mul iple-session condi ions, he pos -p e e ec sizes o he
single-session condi ion we e always a leas as s ong as hose
o he mul iple-session condi ion. In addi ion, he change in
IAT showed a medium e ec size o bo h condi ions, a medium
e ec size o a e SPIC in he single-session condi ion, and a
small e ec size o he o he s. Since pos STAI was only
measu ed jus be o e he conce speech, he e was no pos -p e
e ec size. Mos o he e ec sizes o compa ison be ween he
condi ions we e small o medium, excep o he di e ence
be ween he single-session exposu e and con ol condi ions,
whe e he e ec size was la ge o a e SPIC.
The e ec sizes a e c ude o e all measu es and do no conside
he ela ionship be ween p e and pos measu es, no do hey
include he co a ia e amilia i y (how much pa icipan s we e
amilia wi h he band Di e S ai s). We now conside he
s a is ical model ha included co a ia es.
Table 2. Means (SEs) o esponse a iables by condi ion, pai ed e ec sizes (Cohen d) o di e ences be ween pos and p e measu es, and e ec sizes
(Cohen d) o pos -p e alues compa ing be ween condi ions.
pos STAId
a e SPIC
pos SPICc
PRCA24b
IATa
Condi ion
Single-session exposu e
23.2 (1.27)5.88 (2.94)–2.31 (3.00)5.38 (2.49)0.35 (0.15)Mean (SE)
—e
0.58–0.210.260.62Cohen d pos -p e
Mul iple-session exposu e
20.7 (0.93)–5.36 (6.84)2.29 (4.77)0.00 (3.36)0.30 (0.17)Mean (SE)
—–0.320.160.000.59Cohen d pos -p e
Con ol
20.6 (1.28)–11.73 (3.35)–7.20 (2.31)4.93 (2.50)–0.10 (0.20)Mean (SE)
—–0.50–0.350.24–0.18Cohen d pos -p e
0.5390.577–0.3060.4780.075Cohen d pos -p e (single- s mul iple-session
exposu e)
0.4891.4230.4590.0450.639Cohen d pos -p e (single-session exposu e s
con ol g oup)
0.0260.3170.680–0.4410.564Cohen d pos -p e (mul iple-session exposu e
s con ol g oup)
aIAT: Implici Associa ion Tes .
bPRCA: Pe sonal Repo o Communica ion App ehension.
cSPIC: S a e Pe cei ed Index o Compe ence.
dSTAI: S a e-T ai Anxie y In en o y.
eNo applicable.
S a is ical Analysis o Anxie y-Rela ed Resul s
Summa ies o all he pos e io dis ibu ions o he model’s
pa ame e s a e gi en in Mul imedia Appendix 8, wi h u he
de ails o he esul s including he model’s goodness o i and
he a ailabili y o da a and p og ams o analysis.
The main conce n o in e es was o check whe he he
single-session condi ion is a leas as e ec i e in con ibu ing
o a diminu ion o anxie y compa ed o he mul iple-session
condi ion and whe he hese a e di e en om he con ol g oup.
We conside ed each esponse a iable in u n, no ing ha all
CIs we e subs an ially na owe han hei p io s.
Implici Associa ion Tes
The in e ac ion e ms o mul iple- and single-session condi ions
had CIs mainly in he posi i e egion wi h high (single session,
0.945) and mode a ely high (mul iple sessions, 0.817)
p obabili ies o being posi i e. In o he wo ds, g ea e alues
o p eIAT we e associa ed wi h g ea e le els o pos IAT.
Mo eo e , he 2 CIs we e simila in hei ange. The coe icien
o p eIAT o he con ol g oup showed no e idence o being
di e en om 0 (p obabili y o being posi i e=0.458). These
esul s poin o he e being no o li le di e ence be ween he
mul iple- and single-session exposu es wi h espec o hei
e ec on IAT, and each o hese we e di e en om he con ol
g oup. Hence, pos IAT inc eased wi h p eIAT o he mul iple-
and single-session g oups. Howe e , o he single-session
g oup, he p obabili y ha he a e o inc ease (slope) was g ea e
han 1 was 0.835, whe eas i was 0.601 o he mul iple-session
g oup.
Pe sonal Repo o Communica ion App ehension
In he con ol g oup, he e was clea ly a posi i e slope be ween
pos PRCA24 and p ePRCA24 (p obabili y=1.000, βp c; see
Table S4 in Mul imedia Appendix 8). This slope was educed
in he case o bo h mul iple sessions (p obabili y = 1 – 0.023 =
0.977) and he single session (p obabili y = 1 – 0.077 = 0.923).
The CIs o he in e ac ion e ms o he mul iple- and
single-session g oups we e simila . Al hough bo h mul iple-
and single-session exposu e migh ha e educed he slopes o
pos PRCA24 on p ePRCA24, he e was no impo an di e ence
JMIR Fo m Res 2024 | ol. 8 | e52212 | p. 7h ps:// o ma i e.jmi .o g/2024/1/e52212 (page numbe no o ci a ion pu poses)
Banakou e alJMIR FORMATIVE RESEARCH
XSL
•
FO
Rende X
be ween hem, and hey we e bo h di e en om he con ol
g oup. Hence, bo h single- and mul iple-session in e en ions
educe he p opo ional inc ease in pos PRCA24 wi h espec
o p ePRCA24.
The pos SPIC Va iable
This was adminis e ed p io o he pa icipan s p esen ing he
Di e S ai s welcome speech in on o a i ual audience. In
he con ol g oup, pos SPIC was posi i ely linea ly ela ed o
p eSPIC (p obabili y=1.000, coe icien was posi i e). In he
mul iple-session g oup, he e was a high main e ec (mean
47.78, p obabili y=1.000), bu he slope on p eSPIC was educed
compa ed o he con ol g oup by –0.67 (p obabili y o being
nega i e = 1 – 0.002 = 0.998). The single-session condi ion also
had a high main e ec o 26.29 (p obabili y o being
posi i e=0.895), and he slope on p eSPIC was educed by a
mean o –0.33 (p obabili y o being nega i e = 1 – 0.142 =
0.858). The amilia i y a iable was posi i ely associa ed wi h
pos SPIC (p obabili y=0.810), consis en wi h he likelihood
ha he p esen a ion ask would be less s ess ul o hose wi h
p io knowledge o Di e S ai s.
The a e SPIC Va iable
This was adminis e ed a e he p esen a ion o he Di e S ai s
welcome speech o he audience. As be o e, he con ol condi ion
was posi i ely associa ed wi h p eSPIC (p obabili y=1.000).
The mul iple-session condi ion had a s ong posi i e main e ec
o 68.06 (p obabili y o being posi i e=0.999) bu wi h a
educ ion in he slope on p eSPIC by –1.02 (p obabili y o being
nega i e = 1 – 0.001 = 0.999). The single-session condi ion also
had a s ong posi i e main e ec o 51.42 (p obabili y=0.968)
bu dec eased in slope on p eSPIC by –0.52 (p obabili y = 1 –
0.096 = 0.904). The e was a high p obabili y (0.970) ha
amilia i y wi h Di e S ai s was posi i ely associa ed wi h
a e SPIC. Fo bo h pos SPIC and a e SPIC, he e was a
di e ence o bo h mul iple- and single-session exposu e
compa ed wi h he con ol g oup.
The pos STAI Va iable
The single- and mul iple-session condi ions had simila pos e io
dis ibu ions, and hei main e ec s we e g ea e han hose o
he con ol condi ion. pos STAI was posi i ely associa ed wi h
amilia i y.
F om he s a is ical model, he pos e io dis ibu ions o each
esponse a iable can be ob ained o any alues o he
co a ia es and he amilia i y a iable using equa ion 2. Figu e
3shows all he pos e io dis ibu ions condi ional on he
co a ia es being 10% g ea e han he lowes -possible alue,
he middle alue, and 10% lowe han he highes -possible alue.
Hence, i he ange o he a iable is xmin-xmax and i d = xmax
– xmin, hen hese alues a e xmin + 0.1d, (xmin + xmax)/2,
and xmax – 0.1d, espec i ely. F om Figu e 3, we can see ha
he con ol dis ibu ions we e clea ly di e en om hose o
he mul iple- and single-session condi ions. Mul iple sessions
ga e be e esul s han he single session only in he case o
pos PRCA24, condi ional on p ePRCA24 being a he highes
le el, as de ined ea lie , bu in his case, bo h mul iple- and
single-session condi ions we e supe io in hei e ec s han he
con ol condi ion. Fo pos IAT, pos SPIC, a e SPIC, and
pos STAI, he single-session condi ion was always a leas as
good as he mul iple-session condi ion.
Table 3 shows p obabili ies compu ed om he dis ibu ions in
Figu e 3, pa icula ly he p obabili ies o imp o emen . Fo all
a iables excep pos PRCA24, imp o emen co esponded o
an inc ease in he pos sco e compa ed o he p e sco e. Fo
pos PRCA24, imp o emen co esponded o a dec ease in
sco es. Fo pos STAI, he e was no p e sco e. Fo mally, he
p obabili ies a e as ollows:
P(pos > x | p e = x, obse ed da a)
(7)
whe e x is he lowes and middle o all a iables excep
pos PRCA24.
Fo pos PRCA24, he p obabili ies a e as ollows:
P(pos < x | p e = x, obse ed da a)
(8)
whe e x is he middle and highes . In he case o pos STAI,
he e was no condi ioning on p e=x.
Fo PRCA24, since lowe alues indica ed be e ou comes wi h
espec o public speaking anxie y, we we e only in e es ed in
he p obabili y ha pos PRCA24 (a e exposu e) was less han
ha when p ePRCA24 was equal o he middle and highes
sco es be o e exposu e (ie, hese indica e imp o emen ). Fo
all o he a iables, highe alues ep esen ed be e ou comes,
so we we e in e es ed in he p obabili y ha he pos alues
we e g ea e han he p e alues o he lowes and middle p e
se ings. P obabili y ~ lowes means p obabili y > lowes o
p obabili y < lowes , depending on whe he he symbol abo e
he column o p obabili ies is > o <. Fo example, he
p obabili y ha pos PRCA24<110.4 condi ional on
p ePRCA24=110.4 was 0.958 in he single-session condi ion.
Fo pos STAI, he e we e no p e alues, so he p obabili y ha
pos STAI>0 was 0.510 in he con ol condi ion and 0.900 in
he mul iple-session condi ion. Whe e ele an in equa ion 2,
he amilia i y a iable was se a i s median alue 3.
F om Table 3, we can see ha he con ol condi ion had he
leas p obabili y o imp o emen in e e y case bu 2 (lowes
IAT and middle PRCA24). Gene ally, he single-session
condi ion was only no ably wo se han he mul iple-session
condi ion in he case o he middle pos PRCA24, bu in his
case, he p obabili ies we e in any case low. Fo a e SPIC, he
single-session condi ion had a p obabili y o imp o emen mo e
han 9 imes g ea e han ha o he mul iple-session condi ion
o he middle sco e.
We also conside ed he odds o imp o emen (ie, a ios o
p obabili ies). In he case o pos IAT, wi h he p eIAT a he
middle se ing, he odds o imp o emen we e 0.257/0.001 =
257 o he mul iple-session condi ion compa ed o he con ol
condi ion and 0.631/0.257 = 2.1 o he single-session condi ion
compa ed o he mul iple-session condi ion. Fo pos PRCS24,
wi h he highes p ePRCS24, he odds o imp o emen o he
mul iple- and single-session condi ions o e he con ol
condi ion we e a leas 1.7 (0.958/0.577). Fo pos SPIC, he
odds o imp o emen wi h he middle p eSPIC we e 27 o he
JMIR Fo m Res 2024 | ol. 8 | e52212 | p. 8h ps:// o ma i e.jmi .o g/2024/1/e52212 (page numbe no o ci a ion pu poses)
Banakou e alJMIR FORMATIVE RESEARCH
XSL
•
FO
Rende X
mul iple-session condi ion and 24 o he single-session
condi ion o e he con ol condi ion. Fo lowes a e SPIC, he
mul iple- and single-session condi ions had odds o imp o emen
o 4.2 and 4.3, espec i ely, and in he case o middle a e SPIC,
he mul iple-session condi ion had odds o imp o emen o 53
o e he con ol condi ion, while he single-session condi ion
had odds o imp o emen o 9 o e he mul iple-session
condi ion. Fo middle pos STAI, he odds o imp o emen we e
1.8 and 1.9 o he mul iple- and single-session condi ions,
espec i ely, o e he con ol condi ion.
Figu e 3. Pos e io dis ibu ions o each esponse a iable using equa ion 2. The ed cu es a e o he single-session condi ion, g een o he
mul iple-session condi ion, and blue o he con ol condi ion. Fo all excep pos STAI, hese a e condi ional dis ibu ions wi h he p e a iable: lowes
(L), middle (M), o highes (H) alues. Fo pos STAI, i is he uncondi ional dis ibu ion. The amilia i y a iable is se a i s median o 3. IAT: Implici
Associa ion Tes ; PRCA: Pe sonal Repo o Communica ion App ehension; SPIC: S a e Pe cei ed Index o Compe ence; STAI: S a e-T ai Anxie y
In en o y.
JMIR Fo m Res 2024 | ol. 8 | e52212 | p. 9h ps:// o ma i e.jmi .o g/2024/1/e52212 (page numbe no o ci a ion pu poses)
Banakou e alJMIR FORMATIVE RESEARCH
XSL
•
FO
Rende X
55. Shadinge D, Ka sion J, Myllykangas S, Case D. The impac o a posi i e, sel - alk s a emen on public speaking anxie y.
Coll Teach. Oc 29, 2019;68(1):5-11. [doi: 10.1080/87567555.2019.1680522]
56. Simons DJ, Le in DT. Change blindness. T ends Cogn Sci. Oc 1997;1(7):261-267. [doi: 10.1016/s1364-6613(97)01080-2]
57. Simons DJ, Rensink RA. Change blindness: pas , p esen , and u u e. T ends Cogn Sci. Jan 2005;9(1):16-20. [doi:
10.1016/j. ics.2004.11.006] [Medline: 15639436]
58. Laloyaux C, De ue C, Doyen S, Da id E, Clee emans A. Unde ec ed changes in isible s imuli in luence subsequen
decisions. Conscious Cogn. Sep 2008;17(3):646-656. [FREE Full ex ] [doi: 10.1016/j.concog.2007.03.002] [Medline:
17446088]
59. S einicke F, B ude G, Hin ichs K, Willemsen P. Change blindness phenomena o i ual eali y display sys ems. IEEE
T ans Visual Comp G aphics. Sep 2011;17(9):1223-1233. [doi: 10.1109/ cg.2011.41]
60. Ma in D, Sun X, Gu ie ez D, Masia B. A s udy o change blindness in imme si e en i onmen s. IEEE T ans Visual Comp
G aphics. May 2023;29(5):2446-2455. [doi: 10.1109/ cg.2023.3247102]
61. Senel G, Macia-Va ela F, Gallego J, Jensen HP, Ho nbæk K, Sla e M. Impe cep ible body ans o ma ion in i ual eali y:
saliency o sel ep esen a ion. iScience. Oc 20, 2023;26(10):107938. [FREE Full ex ] [doi: 10.1016/j.isci.2023.107938]
[Medline: 37876610]
Abb e ia ions
CBT: cogni i e beha io al he apy
ET: exposu e he apy
HMD: head-moun ed display
IAT: Implici Associa ion Tes
PRCA: Pe sonal Repo o Communica ion App ehension
PRCS-12: 12-i em Pe sonal Repo o Con idence as a Speake
RCT: andomized con olled ial
SPIC: S a e Pe cei ed Index o Compe ence
STAI: S a e-T ai Anxie y In en o y
VR: i ual eali y
VRCBT: cogni i e beha io al he apy in i ual eali y
VRET: i ual eali y exposu e he apy
Edi ed by A Ma agani; submi ed 26.08.23; pee - e iewed by A Bus illo; commen s o au ho 23.03.24; e ised e sion ecei ed
29.04.24; accep ed 08.05.24; published 22.07.24
Please ci e as:
Banakou D, Johns on T, Beacco A, Senel G, Sla e M
Desensi izing Anxie y Th ough Impe cep ible Change: Feasibili y S udy on a Pa adigm o Single-Session Exposu e The apy o Fea
o Public Speaking
JMIR Fo m Res 2024;8:e52212
URL: h ps:// o ma i e.jmi .o g/2024/1/e52212
doi: 10.2196/52212
PMID: 39037760
©Domna Banakou, Tania Johns on, Alejand o Beacco, Gizem Senel, Mel Sla e . O iginally published in JMIR Fo ma i e Resea ch
(h ps:// o ma i e.jmi .o g), 22.07.2024. This is an open-access a icle dis ibu ed unde he e ms o he C ea i e Commons
A ibu ion License (h ps://c ea i ecommons.o g/licenses/by/4.0/), which pe mi s un es ic ed use, dis ibu ion, and ep oduc ion
in any medium, p o ided he o iginal wo k, i s published in JMIR Fo ma i e Resea ch, is p ope ly ci ed. The comple e
bibliog aphic in o ma ion, a link o he o iginal publica ion on h ps:// o ma i e.jmi .o g, as well as his copy igh and license
in o ma ion mus be included.
JMIR Fo m Res 2024 | ol. 8 | e52212 | p. 16h ps:// o ma i e.jmi .o g/2024/1/e52212 (page numbe no o ci a ion pu poses)
Banakou e alJMIR FORMATIVE RESEARCH
XSL
•
FO
Rende X