educ-07-1007803 No embe 17, 2022 Time: 16:32 # 1
TYPE O iginal Resea ch
PUBLISHED 23 No embe 2022
DOI 10.3389/ educ.2022.1007803
OPEN ACCESS
EDITED BY
Min Lu,
Uni e si y o Miami, Uni ed S a es
REVIEWED BY
Timo hy Adeliyi,
Du ban Uni e si y o Technology,
Sou h A ica
Nim od Delan e,
James Cook Uni e si y Singapo e,
Singapo e
*CORRESPONDENCE
Ma iel F. Musso
[email p o ec ed]
SPECIALTY SECTION
This a icle was submi ed o
Assessmen , Tes ing and Applied
Measu emen ,
a sec ion o he jou nal
F on ie s in Educa ion
RECEIVED 30 July 2022
ACCEPTED 07 No embe 2022
PUBLISHED 23 No embe 2022
CITATION
Boekae s M, Musso MF and
Cascalla EC (2022) P edic ing
a ibu ion o le e w i ing
pe o mance in seconda y school:
A machine lea ning app oach.
F on . Educ. 7:1007803.
doi: 10.3389/ educ.2022.1007803
COPYRIGHT
© 2022 Boekae s, Musso and
Cascalla . 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 (CC BY). The use, dis ibu ion
o ep oduc ion in o he o ums is
pe mi ed, p o ided he o iginal
au ho (s) and he copy igh owne (s)
a e c edi ed and ha he o iginal
publica ion in his jou nal is ci ed, in
acco dance wi h accep ed academic
p ac ice. No use, dis ibu ion o
ep oduc ion is pe mi ed which does
no comply wi h hese e ms.
P edic ing a ibu ion o le e
w i ing pe o mance in
seconda y school: A machine
lea ning app oach
Monique Boekae s1, Ma iel F. Musso2,3*and
Edua do C. Cascalla 4
1Facul y o Social and Beha iou al Sciences, Leiden Uni e si y, Leiden, Ne he lands, 2Depa men o
Expe imen al Psychology, Uni e si y o G anada, G anada, Spain, 3In e disciplina y Cen e o
Resea ch in Ma hema ical and Expe imen al Psychology (CIIPME), Na ional Council o Scien i ic
and Technical Resea ch (CONICET), Buenos Ai es, A gen ina, 4Facul y o Psychology and
Educa ional Sciences, KU Leu en, Leu en, Belgium
The lea ning esea ch li e a u e has iden i ied he complex and
mul idimensional na u e o lea ning asks, in ol ing no only (me a) cogni i e
p ocesses bu also a ec i e, linguis ic, and beha io al con ex ualized aspec s.
The p esen s udy aims o analyze he in e ac ions among ac i a ed domain-
speci ic in o ma ion, con ex -sensi i e app aisals, and emo ions, and hei
impac on ask engagemen as well as ask sa is ac ion and a ibu ion o
he pe cei ed lea ning ou come, using a machine lea ning app oach. Da a
was collec ed om 1130 oca ional high-school s uden s o bo h gende s,
be ween 15 and 20 yea s o age. P ospec i e ques ionnai es we e used o
collec in o ma ion abou he s uden s’ home en i onmen and domain-
speci ic a iables. Mo i a ion p ocesses ac i a ed du ing he lea ning episode
we e measu ed wi h Boekae s’ on-line mo i a ion ques ionnai e. The aces
ha s uden s le behind we e also inspec ed (e.g., ime spen , use o p o ided
ools, con en , and echnical aspec s o w i ing). A i icial neu al ne wo ks
(ANN) we e used o p o ide in o ma ion on he mul iple in e ac ions be ween
he measu ed domain-speci ic a iables, si ua ion-speci ic app aisals and
emo ions, ace da a, and backg ound a iables. ANN could iden i y wi h
high p ecision s uden s who used a w i ing skill, a ec , and sel - egula ion
s a egies a ibu ion on he basis o domain a iables, app aisals, emo ions,
and pe o mance indica o s. ANN de ec ed impo an di e ences in he
ac o s ha seem o unde lie he s uden s’ causal a ibu ions.
KEYWORDS
a ibu ion, app aisals, a i icial neu al ne wo ks, emo ions, lea ning ou come
F on ie s in Educa ion 01 on ie sin.o g
educ-07-1007803 No embe 17, 2022 Time: 16:32 # 2
Boekae s e al. 10.3389/ educ.2022.1007803
In oduc ion
Sel - egula ion plays a key ole in lea ning and s udies ha e
shown ha he le el o sel - egula ion exhibi ed by s uden s
o a ying ages is co ela ed wi h hei lea ning pe o mance
(E klides,2011). The e m sel - egula ed lea ning (SRL) is
made up o h ee in e ela ed componen s, namely he Sel ,
egula ion, and lea ning. These h ee componen s e e o h ee
la ge bodies o knowledge ha oge he desc ibe he cogni i e
p ocesses ha make up lea ning; he me acogni i e knowledge
and s a egies ha egula e he lea ning p ocess, and he sel -
ela ed mo i a ional p ocesses ha con ol he alloca ion o
esou ces ac oss on- ask and o - ask ac i i ies du ing ask
engagemen .
Al hough se e al esea ch g oups p oposed complex,
in eg a ed models o unde s and and esea ch aspec s o SRL
(Boekae s and Niemi i a,2005;Zimme man and Schunk,
2011;Schunk and Zimme man,1998), hey ha e downplayed
he e ec ha cogni ions and emo ions ha e on ask engagemen
and on he s uden s’ pe cep ion and in e p e a ion o he
lea ning ou come (Boekae s and Pek un,2016). The Dual
P ocessing SRL model was an a emp o ill ha gap (e.g.,
Boekae s and Niemi i a,2005). This dynamic and si ua ed,
holis ic amewo k allows esea che s o explo e he in e ac ions
be ween mul iple in e wined aspec s o SRL. Howe e , he
complexi y o he p ocesses in ol ed in SRL needs a mo e
obus and accu a e me hodological app oach which has been
success ul in add essing p oblems in educa ion and he social
sciences (e.g., Fong e al.,2009;Kanakana and Olan ewaju,2011;
Abu Nase ,2012;Boekae s e al.,2012;Musso e al.,2012,
2013,2020;Rod iguez-He nandez e al.,2021). A machine-
lea ning me hod, a i icial neu al ne wo ks (ANN), has been
shown o be e y e ec i e in he s udy o p oblems consis ing o
a la ge numbe o a iables in complex, non-linea , and poo ly
unde s ood in e ac ions (Cascalla e al.,2015). These ANN
a e powe ul classi ie s which build plausible a chi ec u es o
explo e he pa icipa ion o a iables in ol ed in he modeling
o a p oblem (Neal and Wu s ,2001;Whi e and Racine,2001;
De ienne e al.,2003). They accomplish his by cons uc ing
a ma hema ical ela ionship by “lea ning” he pa e ns o all
inpu s om each o he indi idual cases used in aining he
ne wo k, while mo e adi ional app oaches assume a pa icula
o m o ela ionship be ween explana o y and ou come a iables
and hen use a a ie y o i ing p ocedu es o adjus he alues
o he pa ame e s in he model. This me hodological app oach
allowed us o analyze a la ge numbe o ac o s wi hou he
usual pa ame ic cons ain s and helped o unde s and he
e ec o indi idual di e ences in cogni i e and sel - egula ion
a iables on he ac o s being s udied, aking in o accoun all
hei complex in e ac ions.
The c i ical ea u e o Dual P ocessing Sel -Regula ed
Lea ning model a e he non-s op cogni i e app aisals and
hei concomi an emo ions. The model posi s ha cogni i e
app aisals – based on ac i a ed me acogni i e and mo i a ional
belie s in in e ac ion wi h si ua ion-speci ic pe cep ions o cues
in he en i onmen – a e igge ed when s uden s a e con on ed
wi h a lea ning ac i i y. These non-s op app aisals di ec and e-
di ec he ocus o a en ion in he SR sys em. When app aisals
and emo ions a e dominan ly posi i e (e.g., s uden s eel sel -
e icacious, alue he ask), a lea ning in en ion is o med and
a en ion is di ec ed o he lea ning ac i i y i sel : i.e., s uden s
a e hen on he mas e y pa hway, meaning ha wo king memo y
(WM) capaci y as well as (me a) cogni i e s a egies a e used
o imp o e lea ning ou come and hey a e mind ully engaged
in he lea ning ask. On he o he hand, when app aisals
and emo ions wi h espec o he ask a e mainly nega i e
(e.g., s uden s expe ience low con ol, eel anxious, bo ed, o
i i a ed), a en ion is di ec ed away om he ask, because hey
a e mo e conce ned wi h hei well-being han wi h lea ning.
Task-i ele an scena ios a e hen dominan : s uden s a e on he
wellbeing pa hway.Boekae s and Co no (2005) desc ibed wo
di e en sel - egula ion s a egies ha s uden s may use o ge
back o he mas e y pa hway, namely emo ion egula ion and
oli ion. Emo ion egula ion may dampen he emo ional a ousal
and make swi ches o he mas e y pa hway mo e p obable
(see also Boekae s,2011;Boekae s and Pek un,2016), while
oli ional s a egies (also called good wo k habi s), help s uden s
o p o ec hei lea ning in en ion when di icul wo k mus be
comple ed, hus e- ou ing ac i i ies om he well-being o he
mas e y pa hway (Wol e s e al.,2013).
A guing ha mo i a ion is con ex ualized, Boekae s
(1999,2011) cla i ied ha s o ed gene al and domain-speci ic
mo i a ional belie s a e ac i a ed by si ua ional cues and
b ough in o WM whe e hey in luence he s uden s’ cu en
cogni ions and eelings. Fo example, when a s uden is
aced wi h a new ma h ask, p io domain knowledge will
be ac i a ed including a sel -e icacy judgmen and his will
gi e ise o a si ua ion-speci ic compe ence app aisal. In o he
wo ds, Boekae s (1999) di e en ia ed be ween domain-speci ic
a iables (in his s udy: sel -e icacy) and si ua ional app aisals
gene a ed du ing a lea ning episode (in his s udy: compe ence
app aisal). She iden i ied se e al ypes o app aisals, including
alue app aisal, compe ence app aisal, pe cei ed di icul y
app aisal, e o app aisal p e and pos , pe cei ed di icul y
app aisal, and sa is ac ion app aisal. Seege s and Boekae s
(1993) hypo hesized and ound ha si ua ional app aisals
media ed he in luence o domain-speci ic mo i a ional belie s
on s uden s’ emo ional s a e and on hei in ended ask
engagemen a he s a o a ma h ask. C ombach e al. (2003)
subs an ia ed hese indings. They showed ha an iden ical
app aisal model i ed he da a collec ed wi h espec o gende
and di e en school subjec s, and ha his model was s able o e
a 6-mon h pe iod. Hence, esul s wi h he Online Mo i a ion
Ques ionnai e (Boekae s,2002) (OMQ) demons a ed ha he
e o ha s uden s a e p epa ed o in es in a lea ning ac i i y
and hei emo ional s a e a he s a o a lea ning episode a e
la gely dependen on he way hey app aise he ask in si u. In a
simila ein, Boekae s e al. (2003) showed ha domain-speci ic
F on ie s in Educa ion 02 on ie sin.o g
educ-07-1007803 No embe 17, 2022 Time: 16:32 # 3
Boekae s e al. 10.3389/ educ.2022.1007803
( ai -like) mo i a ion belie s – measu ed sepa a ely om he
ask – in e ac ed wi h hei cogni i e app aisals (measu ed
du ing he ask) and join ly a ec ed he s uden s’ a ibu ion o
he pe cei ed lea ning ou come.
Belie s abou he Sel in ela ion o speci ic con en o
subjec -ma e a e mo i a ional in na u e. These mo i a ional
belie s ha e been s udied ex ensi ely, including he s uden s’
basic a i udes and mo i es in ela ion o asks in a speci ic
domain, hei epis emic belie s, sel -e icacy judgmen s,
and ou come expec a ions, hei alue and in e es belie s,
a ibu ions, goal o ien a ion, con ol belie s, and pe cei ed
a ailabili y o social suppo . To da e he e is a as body
o li e a u e documen ing ha domain-speci ic mo i a ional
belie s a e signi ican ly linked o he le el and quali y o
s uden s’ in o ma ion p ocessing and hei me acogni i e
s a egy use (Pin ich,2000;Vole ,2001;Ho e and Pin ich,
2002;E klides,2011;Pek un and Pe y,2014) as well as o hei
use o oli ional s a egies (Boekae s and Co no,2005;Wol e s
e al.,2013).
A ibu ions e e o knowledge s uc u es ha help s uden s
o in e p e he cause o a success ul o unsuccess ul lea ning
ou come. Al hough we ully acknowledge he impo ance o
he h ee causal dimensions ad oca ed by Weine and his
colleagues (see Zimme man and Ki san as,2002;G aham and
Taylo ,2014;Weine ,2018) and we alue he insigh s gained
om s udying abili y and e o a ibu ions as pa o a
s uden ’s a ibu ional s yle, we applauded Zimme man and
Ki san as’(2002)con ex ualized app oach o s udy s uden s’
a ibu ions. Since he u n o he cen u y, esea che s in
Educa ional Psychology ha e placed g ea e emphasis on
s a egy use as a eason o success and ailu e in academics. They
explained ha a s a egy a ibu ion di e s om bo h an abili y
(in e nal, s able, uncon ollable) and e o (in e nal, a iable,
con ollable) a ibu ion in he sense ha i p omp s s uden s
o e lec on he easons why a pe cei ed poo ou come was
no con ollable and on whe he his is seen as a pe sis en
sho coming o as luc ua ing o e ime. Fo example, Chan
and Moo e (2006) cla i ied ha s uden s who explain ailu e
in e ms o low abili y do no seem o be awa e ha hey used
ine ec i e s a egies and ha di e en s a egies may exis o
achie e a be e esul . In a simila ein, s uden s who use
an e o a ibu ion (e.g., I did poo ly because I did no pu
in a lo o e o ) may no be able o di e en ia e be ween
‘some e o ,’ ‘a lo o e o ,’ and ‘g ea e o ’ and hey may
no ha e a clue wha exac ly hey need o do di e en ly nex
ime o imp o e hei pe o mance. By con as , s uden s who
e e o a speci ic s a egy o explain hei esul on a ecen
exam (e.g., ‘I did poo ly because I did no ead he ins uc ion
wice’; o , ‘I did well, because I checked all my answe s wice
be o e handing in he exam’) ealize ha hey can imp o e hei
pe o mance nex ime by choosing mo e e ec i e s a egies, o
by eques ing help in he use o speci ic skills be o e aking he
exam.
Impo an ly, s uden s who use a s a egy a ibu ion o
explain ailu e on an exam do no eel loss o con ol as do
s uden s who use a low abili y a ibu ion o a ‘ he domain is oo
di icul o me’ a ibu ion. Zimme man and Ki san as (2002)
and Zimme man and Clea y (2009) epo ed ha an a ibu ion
o e o ha explici ly in ol ed he use o SR s a egies ha
he s uden s had in ended o use – and hence pe cei ed as
con ollable – is mo e mo i a ing because i helps lea ne s
o sus ain pe cep ions o e icacy and in es e o o succeed
du ing pe iods o subop imal pe o mance. I also p o ec s hem
agains ge ing en wined in nega i e emo ion spi als as well as
om nega i e eac ions om o he s (see also Wol e s e al.,
2013).
Gi en ha Weine ’s a ibu ion heo y was modeled on
he basis o ai -like measu es, i cap u es he egula i ies in
s uden s’ cha ac e is ic causal a ibu ions and is la gely blind o
he in e ac ion o ac i a ed a ibu ional belie s wi h pe cei ed
con ex ual cons ain s and a o dances. Ye , conside able
li e a u e om mains eam psychology has e ealed ha
disc e e emo ions, g ounded in cogni i e app aisals, in luence
many di e en cogni i e p ocesses, such as con en and dep h o
in o ma ion p ocessing, e alua i e judgmen s, causal hinking
and decision making (Kel ne e al.,2013). This leads o he
ine i able conclusion ha adi ional a ibu ion esea ch does
no p o ide su icien in o ma ion on he impac o si ua ional
app aisals and emo ions on s uden s’ causal explana ion o he
lea ning ou come.
In his espec i was in e es ing o no e in a ecen e iew o
Weine ’s a ibu ion amewo k ha G aham (2020) desc ibed
he a ibu ion p ocess as a sequence ha begins wi h a pe cei ed
ou come which is in e p e ed as a success o a ailu e. These
cogni ions may igge ou come-dependen emo ions, such as
happiness, sadness, p ide, shame, eg e , o guil . Ou way o
concep ualizing he a ibu ion p ocess is somewha di e en .
Based on he Dual P ocessing SR model, we p opose ha he
a ibu ion p ocess begins as soon as s uden s a e aced wi h a
lea ning ask. Mo e speci ically, hei pe cep ion o he lea ning
ou come and i s concomi an causal a ibu ion begins o ake
shape du ing he lea ning episode and is a ec ed by he mul iple
cogni i e app aisals and emo ions – and hei in e ac ions – ha
ake place while he lea ning episode is un olding.
The p esen s udy
Resea ch ques ions
Two main se s o ques ions we e aised a he s a o he
s udy. Fi s , how does he quali y o he in e nal en i onmen
ha s uden s c ea e in he goal se ing and s i ing s age,
in luence hei causal asc ip ion o he pe cei ed w i ing
ou come? The second ques ion pe ains o he p edic ion
and de e mina ion o he speci ic con ibu ion o mo i a ional
a iables, a he domain le el and a he si ua ion-speci ic le el,
o s uden s’ causal asc ip ions. The hi d ques ion is: can we
p edic wi h high p ecision who will use/will no use a speci ic
a ibu ion using neu al ne wo k (ANN) models?
F on ie s in Educa ion 03 on ie sin.o g
educ-07-1007803 No embe 17, 2022 Time: 16:32 # 4
Boekae s e al. 10.3389/ educ.2022.1007803
We hypo hesized ha s uden s who pe cei e hei w i ing
ou come as ‘good’ a e mo e likely o endo se i ems ha desc ibe
speci ic w i ing skill (WS) as an explana ion o hei sel -
pe cei ed success ul pe o mance han s uden s who pe cei e
hei w i ing ou come as ‘poo .’ We also hypo hesized ha he
s uden s’ sco es on he me acogni i e and mo i a ion belie s
scales – measu ed independen ly om he w i ing episode –
would in e ac wi h hei app aisals and emo ions du ing he
lea ning episode and would make a signi ican con ibu ion
in he p edic ion o hei a ibu ions. Mo e speci ically, ou
hypo heses we e:
(1) S uden s who pe cei e hei w i ing ou come as poo ( he
poo g oup) will endo se less WS a ibu ion i ems o
explain hei w i ing ou come compa ed o s uden s who
pe cei e hei w i ing ou come as good ( he good g oup).
(2) The poo g oup will also endo se less SR s a egy
a ibu ions han he good g oup.
(3) The poo g oup will endo se mo e a ec ela ed
a ibu ions han he good g oup.
(4) The poo g oup will sco e signi ican ly lowe on epo ed
domain-speci ic belie s (sel -e icacy, me acogni i e skills,
epis emic belie s, and will be mo e dependen on ex e nal
egula ion) han he good g oup.
(5) Compa ed wi h he good g oup, he poo g oup will
sco e signi ican ly lowe on compe ence, alue, and e o
app aisal when s a ing he w i ing ask. Also, hei
sco es will be signi ican ly lowe on e o , di icul y,
and sa is ac ion app aisal a e doing he ask han hose
o he good g oup.
(6) The poo g oup will display signi ican ly lowe sco es han
he good g oup on posi i e emo ions and highe sco es on
nega i e emo ions, bo h p e and pos le e w i ing.
(7) Bo h domain a iables and con ex -sensi i e app aisals
and emo ions will con ibu e signi ican ly o he p edic ion
o oca ional s uden s’ a ibu ion o he pe cei ed
w i ing ou come.
(8) Neu al ne wo k (ANN) classi ica o y analysis can iden i y
wi h a high deg ee o accu acy s uden s who endo se/do
no endo se a speci ic a ibu ion on he basis o he
p edic i e weigh s assigned o he inpu s.
(9) The pa e n o hese weigh s will expose he p ima y
indica o s o he quali y o he in e nal model (QIM)
ha lead o an a ibu ion o adequa e/inadequa e
WS, adequa e/inadequa e sel - egula ion s a egies, and
expe ienced posi i e o nega i e a ec .
Measu ing he quali y o he in e nal
a ec i e model
In o de o s udy he quali y o he in e nal en i onmen
ha s uden s c ea e du ing he lea ning episode we modeled
he nume ous ac o s ha cons i u e he s uden s’ men al
amewo k du ing a lea ning episode. Based on he Dual
P ocessing Sel -Regula ion Model, we hypo hesized ha
s uden s ha e access o and ac i a e ele an domain
in o ma ion om long- e m memo y (e.g., he mo i a ion
and me acogni i e belie s and s a egies desc ibed in sec ion
“Ma e ials and me hods”) and b ing i in o WM. This ac i a ed
in o ma ion in e ac s wi h he s uden s’ non-s op cogni i e
app aisals o he ask and he concomi an emo ions (see
Theo e ical F amewo k sec ion). Based on hese in e ac ions he
s uden s make (un)conscious decisions abou hei expec ancies,
a ge s, and he ype o engagemen ha hey a e willing o
commi hemsel es o. They also judge hei p og ess, e alua e
he lea ning ou come, and make a ibu ional judgmen s.
Measu ing he s uden s’ a ibu ions in
a con ex -sensi i e way
D awing on Zimme man’s empi ical indings we easoned
ha he a ibu ion i ems should no only be con ex ualized;
he i ems should also be o mula ed in he s uden s’ own wo ds
so ha hey would uly unde s and wha hey we e asked o
endo se and pe cei e he a ibu ion i ems as ele an o he
cu en lea ning ac i i y. This would encou age hem o e lec
on he conc e e and unc ional aspec s o he a he abs ac
e o and abili y causal judgmen s p oposed by Weine . Bo h
e o and abili y a e in ac umb ella e ms ha lea e a lo o
oom o be in e p e ed in a ques ionnai e. Fo example, e o
may be in e p e ed as a sign o in ol emen and engagemen ,
bu also as a sign o low abili y in esponse o pe cei ed
di icul y, low sel -e icacy, and anxie y. Hence, we e ised he
a ibu ion i ems o he OMQ in e ms o he conc e e skills
and SR s a egies ha he s uden s had supposedly mas e ed
in he class oom (acco ding o he eache g oup [Tg oup]),
and he e o e could d aw on du ing he lea ning episode.
Fo example, we heo ized ha s uden s who endo sed he
a ibu ion i em “I did poo ly on his w i ing ask, because I
canno o mula e co ec sen ences in Du ch” communica e h ee
hings. Fi s , ha hey a e awa e ha his skill is ele an and
impo an o do he ask, second ha hey cu en ly (s ill) lack
ha skill, and hi d ha i is possible o acqui e ha skill.
Ma e ials and me hods
Pa icipan s
Pa icipan s in his esea ch we e 1,130 seconda y oca ional
school s uden s in ou schools ac oss he Ne he lands
(Females = 65.9%). The s uden s’ age anged om 15 o 22 yea s
old. Seconda y oca ional educa ion is a 4-yea oca ionally
o ien a ed (high school) s eam p o iding a p ac ical o ien a ion
F on ie s in Educa ion 04 on ie sin.o g
educ-07-1007803 No embe 17, 2022 Time: 16:32 # 5
Boekae s e al. 10.3389/ educ.2022.1007803
o highe oca ional s udies o o oca ional aining. S uden s
ollowed di e en s udy p og ams, including comme ce, ca e
and wel a e, nu sing, adminis a ion, and educa ional assis an .
As pa o an obliga o y Du ch language cou se, hey had
ecei ed aining in w i ing o communica ion. A week be o e
he ac ual s udy began; s uden s we e old by hei na i e
language eache ha hei class would pa icipa e in a na ion-
wide w i ing session o assess he quali y o hei Du ch WSs
on-line. All s uden s we e na i e speake s and ag eed ha hei
w i ing p oduc s would be sen au oma ically o he se e o
he uni e si y, whe e esea che s would examine he quali y o
he w i ing p ocess and p oduc . They we e in o med, howe e ,
ha his was no a high-s ake na ional assessmen , ha he da a
would be analyzed anonymously, bu ha eedback would be
p o ided o he pa icipa ing eache s a he class le el so ha
hey could imp o e he w i ing pe o mance o hei s uden s.
Ins umen s and measu es
On ad ice om a eache g oup ha helped us o se up
and moni o he s udy (TG oup) a speci ic au hen ic lea ning
ac i i y was selec ed o wo easons. Fi s , s uden s in oca ional
school had been ained o w i e le e s, such as an applica ion
le e o a holiday job, as pa o he cu iculum and hey
also had aining in he use o me acogni i e s a egies o
egula e hei own lea ning. Hence, we hypo hesized ha he
s uden s would be able o men ally ep esen a w i ing ac i i y
in WM in e ms o i s con ollabili y and alue (compe ence
and alue app aisal) and would ha e a good idea abou he
e o in ol ed o ob ain a good w i ing ou come. Second, a
le e w i ing ask has a clea beginning and end so ha he
OMQ could be embedded a ound he le e w i ing ac i i y. This
would allow us o obse e he un olding lea ning p ocess, eco d
hei app aisals and emo ions, and examine ace da a. Because
s uden s’ app aisals and emo ions may change du ing he
w i ing episode due o inc easing o dec easing ask demands
o dis u bing p ocessing bu dens, he OMQ measu es app aisals
and emo ions in he beginning o he lea ning episode (he e:
a e b ie ly inspec ing he ask and lis ening o he ins uc ions),
and again a he end o he lea ning ac i i y (he e: jus be o e
epo ing on he quali y o he w i ing ou come and endo sing
he di e en causal a ibu ions).
Fou di e en ypes o a iables we e measu ed in his s udy,
namely (1) backg ound a iables, (2) mo i a ion belie s and
me acogni i e s a egies, (3) on-line app aisal and emo ions,
and (4) pe o mance indica o s.
Backg ound a iables
They included gende , highes educa ion o mo he / a he ,
pe cei ed quali y o mo he ’s and a he ’s WSs, numbe o
iends in school, daily w i ing habi s ( equency o using cell
phone, emails o iends, dai y use), eeling com o able in class,
and equency doing homewo k wi h iends). These a iables
we e measu ed wi h 5-poin Like - ype scales.
Mo i a ion belie s and me acogni i e s a egies
They we e measu ed wi h con en ional ai - ype scales,
independen ly o he le e w i ing session. The scales cap u ed
(1) s uden s’ sel -e icacy judgmen , (2) goal o ien a ion, (3)
epis emological belie s, (4) access o and use o me a-cogni i e
s a egies, and (5) (in) dependence on ex e nal egula ion. Since
mos o hese scales a e con ex -insensi i e, which lea es us
blind as o how he s uden s in e p e he i ems, we encou aged
s uden s o key in o amilia clus e s o lea ning ac i i ies hey
had pa icipa ed in as pa o he Du ch language cou se, such as
eading and ex p ocessing lessons, w i ing lessons, and Du ch
co espondence. They had o answe se s o 5-poin Like
i ems on-line ocusing on hese amilia pa s o he cou se.
The i ems we e ansla ed om exis ing ques ionnai es. Sel -
e icacy was assessed wi h he alida ed sel -e icacy scale o he
MSLQ (Pin ich and De G oo ,1990). The scale consis s o
Like i ems wi h ancho poin s 1 = o ally disag ee, 5 = o ally
ag ee (a high sco e means ag eemen ha he a iable is
p esen ). An example i em is ‘Compa ed o o he s in his class, I
expec o do well.’ Goal o ien a ion and Me acogni i e s a egies
we e assessed wi h he Goal O ien a ion and Lea ning S a egy
Su ey (GOALS-S), de eloped and alida ed by Dowson and
McIne ney (2004). All i ems we e 5-poin Like scales wi h
1 = s ongly disag ee o 5 = s ongly ag ee. The i ems ha
assessed goal o ien a ion spli up in o mas e y goal o ien a ion
(e.g., I wan o unde s and wha I lea n a school), pe o mance
o ien a ion (e.g., I wan o imp ess o he s wi h my pe o mance),
well-being goal o ien a ion (e.g., I wan o ge on wi h my
classma es) and ca ee o ien a ion goals (e.g., I wan o ge a
good job). Me acogni i e s a egies included elabo a ion (e.g., I
y o unde s and how he hings I lea n in school i oge he ),
planning (e.g., I always hink be o ehand wha I wan o say),
and moni o ing (e.g., I o en s op o see wha pa o my
wo k is no sa is ac o y ye ). Dependency on Ex e nal egula ion
was measu ed wi h a subscale o Ve mun ’s (1998) alida ed
In en o y o Lea ning S yles (ILS). An example i em is: I I do
no unde s and he ins uc ions, I wai o he eache o explain
wha I ha e o do). Epis emological belie s we e measu ed wi h
Schomme ’s (1990) alida ed es . An example i em is: ‘I hink
ha knowledge is s able.’ A high sco e on his scale implies ha
he s uden s iew knowledge as ce ain, ixed, and objec i e.
Cogni i e app aisals and emo ions
A e p esen ing he w i ing ask and i s ins uc ions bu
be o e s a ing on he ask se e al app aisals we e measu ed wi h
Boekae s’ on-line mo i a ion ques ionnai e (OMQ: Boekae s,
2002). This ine-g ained measu emen echnique was alida ed
by Boekae s and he colleagues (see sec ion “In oduc ion”)
and includes: compe ence app aisal (n = 4; alpha = 0.84; e.g., ‘I
can do his w i ing ask well’), alue app aisal ( ask a ac ion,
F on ie s in Educa ion 05 on ie sin.o g

educ-07-1007803 No embe 17, 2022 Time: 16:32 # 6
Boekae s e al. 10.3389/ educ.2022.1007803
n = 1; e.g., ‘ his w i ing ask is in e es ing’ and pe cei ed u ili y,
n=3; alpha = 0.79; e.g., ‘I conside his w i ing ask ele an
o me’), and e o app aisal p e (n= 3; alpha = 0.82; e.g., ‘I
will do he w i ing ask a en i ely’). A he same ime hei
posi i e emo ions p e (N= 3; alpha = 0.83; e.g., I eel good, a
ease, pleasan ) and nega i e emo ions p e (n= 3; alpha = 0.69;
e.g., I eel bo ed, i i a ed, i ed) as well as anxie y p e (n= 1; I
eel ne ous) we e assessed. A ask comple ion, s uden s again
epo ed on he same emo ions [posi i e emo ions pos (n= 3;
alpha = 0.87); nega i e emo ions pos (n= 3; alpha = 0.74) and
anxie y pos (n= 1), e o app aisal pos (n= 1; e.g., ‘I w o e his
le e a en i ely’), di icul y le el app aisal (n= 1, e.g., ‘I ound
i ha d o w i e his le e ’), sa is ac ion app aisal (n= 1; e.g., ‘I
eel sa is ied abou he w i ing ac i i y’), and ou come app aisal
(n = 1; e.g., ‘Wha is he quali y o he le e you w o e?’: e y
poo , poo , good o e y good)]. Finally, hey causally a ibu ed
he pe cei ed w i ing ou come (e.g., I did poo ly because I did
no know wha o say in he le e /I did well because I can
s uc u e a le e well). A h ee-s ep p ocedu e was used o
design he con ex ualized a ibu ion i ems. We i s asked he
TG oup o lis he possible easons ha hei s uden s gi e o a
poo and good w i ing pe o mance, ocusing especially on hei
use o WSs and sel - egula ion s a egies. Second, we es ed hese
o mula ions wi h s uden s in oca ional school, asking hem o
help us e o mula e he i ems in hei own wo ds and ex end
he lis i hey hough ha impo an easons o explaining
success and ailu e in lea ning ou comes we e missing. Thi d,
we ex ended Boekae s’ OMQ wi h hese i ems, inco po a ing
hem in o he so wa e p og am. S uden s who pe cei ed hei
lea ning ou come as ‘good’ e lec ed on whe he hey ag eed o
disag eed ha access and use o speci ic WSs, sel - egula ion
s a egies, and/o a ec - ela ed a iables we e he eason why
hey did well on he ask. By con as , s uden s who pe cei ed
hei w i ing ou come as ‘poo ’ judged whe he hey ag eed
o disag eed ha lack o speci ic WS, ailu e o use e ec i e
Sel - egula ion, and/o a ec - ela ed a iables we e he eason
why hey did no do well on he ask. All app aisals, emo ions,
and a ibu ions we e measu ed on 5-poin Like scales wi h
1 = s ongly disag ee o 5 = s ongly ag ee. Resul s o he
explo a o y ac o analyses (EFA) o he a ibu ion i ems ca ied
ou in he p esen sample, esul ed in a h ee- ac o model
o bo h ypes o ou come app aisals (poo and good w i ing):
WSs (α= 0.82/0.85, espec i ely); Sel - egula ion S a egies
(α= 0.65/0.74), and Expe ienced A ec (α= 0.60/0.66). This
s uc u e was con i med by a con i ma o y ac o analysis (CFA)
p o iding cons uc alidi y o hese ad-hoc a ibu ion scales
( o good ou come: CFI = 0.895; NFI = 0.883; RMSEA = 0.09;
o poo ou come: CFI = 0.968; NFI = 0.965; RMSEA = 0.09).
The in e nal consis ency o he h ee scales was sa is ac o y in
he good as well as in he poo g oup, e en hough he e we e
ew i ems in he A ec and Sel - egula ion S a egies1.
Pe o mance indica o s
They e e o he aces le behind by he s uden s: he ime
spen w i ing he le e was eco ded au oma ically in seconds
(LAG) and an au oma ic coun o he numbe o imes s uden s
pushed any o he ool bu ons p o ided he ool use sco e. Tools
we e a ailable o wo d coun , dic iona y use, spelling checke ,
ecall o ins uc ions, ecall o he newspape a icle, eques o
model example o a le e o he edi o (N= 6; alpha = 0.71).
In o de o es ablish le els o w i ing pe o mance, wo Ango
s anda d se ing sessions we e ca ied ou (Cizek,2006). Fi s ,
Technical Sco e s anda ds we e es ablished, ollowed by he
Con en Sco e s anda ds. Nex , pe o mance le el desc ip ions
we e w i en o highly compe en / compe en / and ma ginally
compe en le e w i e s, sepa a ely o echnical w i ing and
con en aspec s and a sco ing key was designed, speci ying wha
should be sco ed as well as he a ached weigh s. The con en
sco e was based on whe he s uden s di ec ed he le e a an
audience, w o e a sho in oduc ion, e e ed explici ly o he
a icle in he newspape , men ioned he oile sea , glue on he
sea , compensa ion money, and ga e hei opinion on whe he
compensa ion should be gi en (conclusion). The numbe o
a gumen s used and he co ec ness o he a gumen s, as
well as he consis ency in he di ec ion o w i ing we e also
sco ed. The echnical sco e was based on equency o e o s
in spelling, conjuga ion, capi aliza ion, punc ua ion, and choice
o w ong wo ds. Two eache s olun ee ed o comple e he
sco ing shee o all he le e s. They we e eed om eaching
esponsibili ies and ained in using he sco ing key un il he
in e - a e eliabili y was sa is ac o y (>0.80). Technical aspec s
we e summa ized in Z-sco e Technical and con en aspec in
he Z-sco e Con en . A subse o s uden p oduc ions (25%)
was double a ed by bo h eache s in o de o check o e all
in e a e ag eemen (Con en : Cohen’s κ= 0.69; Technical:
Cohen’s κ= 0.75). Howe e , he o iginal sco e was main ained
o he analyses.
P ocedu e
E hical app o al was ob ained om he Resea ch E hics
Commi ee o he Leiden Uni e si y. The ea men o
1 The ecological alidi y o ou a ibu ion measu es is high because we
used a clea ly de ined lea ning ac i i y wi h a de ini e beginning and end.
The ine-g ained ques ions ha we designed we e explici ly linked o he
lea ning ac i i y, hus educing he ambigui y and agueness ha is so
ypical o de-con ex ualized ypes o a ibu ion i ems. As Clea y (2011)
con incingly showed he use o con ex and ask-speci ic ques ions ha e
he po en ial o gene a e highly eliable and alid in o ma ion, mainly
because he i ems measu e hough s, belie s, and eelings ha occu in
eal ime, he e o e minimizing he e ec s o memo y o o he e ie al
p oblems (p. 338).
F on ie s in Educa ion 06 on ie sin.o g
educ-07-1007803 No embe 17, 2022 Time: 16:32 # 7
Boekae s e al. 10.3389/ educ.2022.1007803
pa icipan s was in acco dance wi h es ablished e hical
s anda ds o he E hical P inciples o Psychologis s and Code o
Conduc (Ame ican Psychological Associa ion,2017). Pa en s
we e in o med abou he pu pose o he s udy, and hey ga e
w i en consen o pa icipa e in he s udy.
S uden s ook pa in h ee sepa a e sessions, which ook
place in he compu e oom in hei own school. Two o
hese sessions a e impo an o he p esen s udy, namely he
ques ionnai e session and he le e w i ing session. Each o
hese sessions las ed a maximum o 50 min. S uden s logged on
o he In e ne si e o ou esea ch g oup and hei esponses
and ac i i ies we e au oma ically eco ded. They could log o
as soon as hey we e eady wi h he ques ionnai es o wi h he
le e and we e allowed o engage quie ly in o he ac i i ies in he
same oom. The ques ionnai e session consis ed o comple ing
se e al on-line ques ionnai es, including mo i a ion belie s and
me acogni i e s a egies, and backg ound a iables. In he le e
w i ing session s uden s we e p esen ed on-line wi h a sho
a icle ha had appea ed in a local newspape a ew weeks be o e
he s udy began (127 wo ds). They we e eques ed o ead he
s o y and w i e a le e o he edi o o a local newspape in he
space p o ided, explaining in hei own wo ds why hey hough
he main cha ac e in he s o y was (o was no ) eligible o
inancial compensa ion. They we e in o med ha hey should
aim o app oxima ely 100 wo ds. The p o ided ool bu ons
we e also explained, and hey we e encou aged o use hem. The
equency o he s uden s’ use o he di e en ools was eco ded
au oma ically. As soon as he w i ing ask was in oduced, on-
line ask p obes appea ed on sc een and p omp ly a ed he
s uden s’ app aisals and emo ions. A e inishing he le e ,
s uden s app aised he esul o he le e w i ing ac i i y as
e y poo o poo (poo lea ning ou come app aisal) o as
good/ e y good (good lea ning ou come app aisal). A poo
ou come app aisal immedia ely p oduced a se o nega i ely
o mula ed a ibu ion i ems on sc een (e.g., ‘I did poo ly on
his ask, because I canno o mula e co ec sen ences’). A good
ou come app aisal p oduced he equi alen se o posi i ely
o mula ed i ems (e.g., ‘I did well on his ask, because I can
o mula e co ec sen ences’).
Analyses
One o he aims o ou s udy was o explo e how
a ibu ions ake shape du ing he ac ual w i ing p ocess. Mo e
speci ically, we easoned ha he QIM ha s uden s c ea e
du ing he w i ing p ocess may di e undamen ally and ha
his quali y in luences he ype o a ibu ions hey make. We
measu ed ele an backg ound a iables, mo i a ion belie s and
me acogni ion pe aining o he language domain as well as
si ua ion-speci ic a iables and ob ained ace da a. As se e al
esea che s ha e poin ed ou (see Cascalla e al.,2006) mo e
adi ional echniques such as mul iple eg ession lose a lo o
in o ma ion due o mul i-co-linea i y and missing da a.
The me hod chosen o he analysis was a mul ilaye
pe cep on neu al ne wo k wi h a backp opaga ion algo i hm.
The aim o hese ANN models is o maximize classi ica ion
accu acy and wo k h ough he da a in an i e a i e p ocess
un il maximum accu acy is achie ed, au oma ically modeling all
in e ac ions among a iables (49 in ou s udy). They a e e y
obus gene al unc ion es ima o s consis ing o a mul ilaye
ne wo k composed o nonlinea uni s, which compu es i s
ac i a ion le el by summing all he weigh ed ac i a ions i
ecei es and which hen ans o ms i s ac i a ion in o a
esponse ia a nonlinea ans e unc ion, which es ablishes
a ela ionship be ween he inpu s and he weigh s hey a e
assigned. A sys ema ic amewo k p oposed by Alyahyan
and Dü¸s egö (2020) was ollowed in o de o implemen
and e alua e he models. Two de elopmen phases o each
p edic i e classi ica ion model we e ca ied ou : aining o
he ne wo k and es ing o he ne wo k ha had been
de eloped in he es ing phase. Du ing he aining phase,
he sys em e alua es he e ec o he weigh pa e ns on
he p ecision o hei classi ica ion o ou pu s, and hen,
h ough backp opaga ion, i adjus s hose weigh s in a ecu si e
ashion. I makes adjus men s o he weigh s o he in e -
co ela ions be ween he inpu a iables in o de o dec ease
he e o wi h he known ou comes. In o ma ion is passed back
h ough he ne wo k in many i e a ions, g adually changing
he weigh s. As aining p og esses, he ne wo k becomes
inc easingly accu a e in eplica ing he known ou comes. This
p ocess is epea ed many imes, and he ne wo k con inues
o imp o e i s p edic ions un il one o mo e o he s opping
c i e ia ha e been me . A minimum le el o accu acy can be
se as s opping c i e ion, al hough addi ional s opping c i e ia
may be used as well (e.g., numbe o i e a ions, amoun o
ime) (Musso e al.,2012,2013). Du ing his aining phase
se e al models we e a emp ed, and se e al modi ica ions o
he neu al ne wo k hype pa ame e s we e explo ed, such as:
lea ning pe sis ence (con olling he con inua ion o aining
a e no signi ican change in weigh s), lea ning a e ( he
a e a which he ANN “lea ns” by con olling he size o
weigh and bias changes du ing lea ning), momen um (adds
a ac ion o he p e ious weigh upda e o he cu en one,
and is used o p e en he sys em om con e ging o a local
minimum), numbe o hidden laye s, s opping ules (when
he ne wo k should s op “lea ning” o a oid o e - i ing he
cu en sample), ac i a ion unc ions (which de ine he ou pu
o a node gi en an inpu o se o inpu s o ha node o
uni ), and numbe o nodes. This modeling p ocess con inued
un il he desi ed le els o accu acy in he classi ica ions we e
achie ed, maximizing he accu acy o he inal model. Once
ained, he ne wo k can be applied, wi h i s s uc u e and
pa ame e s, o u u e cases ( alida ion o holdou sample)
F on ie s in Educa ion 07 on ie sin.o g
educ-07-1007803 No embe 17, 2022 Time: 16:32 # 8
Boekae s e al. 10.3389/ educ.2022.1007803
TABLE 1 A chi ec u e o NN models.
Pe cei ed poo pe o mance Pe cei ed good pe o mance
Measu es NN1:
Low WS
a ibu ion/
high
NN2:
Low a ec
a ibu ion/
high
NN3:
Low SRS
a ibu ion/
high
NN4:
High WS
a ibu ion/
low
NN5:
High a ec
a ibu ion/
low
NN6:
High SRS
a ibu ion/
low
C oss-en opy e o
S opping e o
0.055
Maximum numbe
o epochs: 1209
0.025
A s opping ule o 1
consecu i e s ep wi h
no dec ease in e o .
0.004
A s opping ule o 1
consecu i e s ep wi h
no dec ease in e o
0.889
A s opping ule o 1
consecu i e s ep wi h
no dec ease in e o
2.695
A s opping ule o 1
consecu i e s ep wi h
no dec ease in e o
3.517
A s opping ule o 1
consecu i e s ep wi h
no dec ease in e o
Lea ning epochs 10 10 10 10 10 10
S opping c i e ion Maximum aining
ime: 15 min
Maximum aining
ime: 15 min
Maximum aining
ime: 15 min
Maximum aining
ime: 15 min
Maximum aining
ime: 15 min
Maximum aining
ime: 15 min
Numbe o ac o s 12 12 12 12 12 12
Numbe o co a ia es 40 40 40 40 40 40
Me hod o escaling
co a ia es
S anda dized
me hod
S anda dized
me hod
S anda dized
me hod
S anda dized
me hod
S anda dized
me hod
S anda dized
me hod
Numbe o hidden laye 1 hidden laye wi h 7
uni s
1 hidden laye wi h 9
uni s
1 hidden laye wi h 8
uni s
1 hidden laye wi h
11 uni s
1 hidden laye wi h
10 uni s
1 hidden laye wi h 6
uni s
Ac i a ion unc ion o
hidden laye s
Hype bolic angen Hype bolic angen Hype bolic angen Hype bolic angen Hype bolic angen Hype bolic angen
Ou pu laye 2 uni s 2 uni s 2 uni s 2 uni s 2 uni s 2 uni s
Ac i a ion and e o unc ion
o ou pu laye
So max, and he
e o unc ion he
c oss-en opy
So max, and he
e o unc ion he
c oss-en opy.
So max, and he
e o unc ion he
c oss-en opy.
So max, and he
e o unc ion he
c oss-en opy.
So max, and he
e o unc ion he
c oss-en opy.
So max, and he
e o unc ion he
c oss-en opy.
o u he alida ion s udies and p og am implemen a ion
(Lippmann,1987). As long as he basic assump ions o he
popula ion o pe sons o e en s ha he ANN used o
aining is cons an o a ies sligh ly and/o g adually, i
can adap and imp o e i s pa e n ecogni ion algo i hms
he mo e da a i is exposed o in he implemen a ions. The
a chi ec u e o ANN models used in his s udy a e p esen ed in
Table 1.
O he machine lea ning algo i hms (i.e., andom o es ,
naï e Bayes, and g adien boos ed ees) as well as mo e
adi ional s a is ical echniques (i.e., gene al linea model and
logis ic eg ession) we e implemen ed using RapidMine S udio
Educa ional 9.10.011. Thus, he e alua ion me ics o he ANNs
could be compa ed agains he o he p edic i e me hodologies.
Resul s
A ibu ions
A o al o 886 s uden s assessed hei w i ing pe o mance as
‘good’ (good g oup), and 244 assessed i as ‘poo ’ (poo g oup).
Gi en s uden s we e dis ibu ed in o di e en g oups acco ding
o hei pe cei ed w i ing ou come and a ibu ion, p incipal
componen analyses we e pe o med, sepa a ely o he poo
and he good g oup. As can be seen in Tables 2,3, h ee main
ac o s eme ged in he good as well as in he poo g oup. The
i s componen g ouped i ems ha explain w i ing ou come
in e ms o echnical WSs, such as ‘I did well (poo ly) because
I can (no ) o mula e co ec sen ences’ and ‘. . .because I (do
no ) know how o s uc u e a le e .’ Hence, his componen
g oups he six i ems ha e e o he s uden s’ w i ing abili y
TABLE 2 P incipal componen analysis o he a ibu ion i ems o
s uden s who pe cei ed hei w i ing ou come as ‘poo ’.
I did poo ly on his ask Componen
1 2 3
•Because I canno o mula e co ec sen ences 0.815
•Because I am no amilia wi h w i ing le e s 0.793
•Because I do no know how o s uc u e a le e 0.767
Because I am poo a w i ing up wha I wan o
say
0.748
•Because I did no know wha o say in my le e 0.672
•Because I could no ind he igh wo ds 0.653
•Because I did no check whe he e e y hing I
wan ed o say was in my le e
0.811
•Because I did no make use o he p o ided ools 0.806
•Because I did no check whe he my le e i ed
he ins uc ions
0.797
•Because I do no like w i ing le e s 0.863
•Because he ask was no in e es ing 0.772
Ro a ion me hod: Va imax wi h Kaise no maliza ion.
F on ie s in Educa ion 08 on ie sin.o g
educ-07-1007803 No embe 17, 2022 Time: 16:32 # 9
Boekae s e al. 10.3389/ educ.2022.1007803
TABLE 3 P incipal componen analysis o he a ibu ion i ems o
s uden s who pe cei ed hei w i ing ou come as ‘good’.
I did well on his ask Componen
1 2 3
•Because I can o mula e co ec sen ences 0.808
•Because I know how o s uc u e a le e 0.785
•Because I am amilia wi h w i ing le e s 0.706
Because I could ind he igh wo ds 0.691
•Because I am good a w i ing up <ha I wan o
say
0.681
•Because I knew wha o say in my le e 0.559
•Because I checked whe he e e y hing, I
wan ed o say was in my le e
0.754
•Because I made use o he p o ided ools 0.749
•Because I checked whe he my le e i ed he
ins uc ions
0.697
•Because he ask was in e es ing 0.850
•Because I like w i ing le e s 0.829
Ro a ion me hod: Va imax wi h Kaise no maliza ion.
in e ms o he speci ic ules and egula ions ha a e needed
o w i e a good le e . This componen was labeled WS. The
second componen consis o he h ee i ems ha e e o
he use o sel - egula ion s a egies lea ned in class (SRS) o
s uc u e one’s own lea ning p ocess, such as ‘. . . because I
checked (did no check) he ins uc ions’ and ‘. . . because I
made / did no make use o he ools p o ided.’ Finally, he
hi d componen is made up o he wo i ems ha e e o
he a ec i e (o mood) s a e he s uden was in du ing he
ask, o example ‘. . .. because he ask was (no ) in e es ing’
and ‘. . . because I like (do no like) w i ing le e s.’ This
componen was labeled ‘Expe ienced A ec ’ (A ). The in e -
co ela ions be ween he h ee a ibu ion componen s we e
low o modes , excep o he co ela ion be ween Componen
1 and Componen 2, which was mode a ely s ong o he
good and he poo g oup ( = 0.43/0.36 espec i ely). This
inding sugges s ha s uden s who endo se a WS a ibu ion
may also use an SRS a ibu ion o explain hei w i ing
ou come. The o me a ibu ion may be mo e s able han
he la e , because i g oups i ems ha ep esen conc e e and
unc ional WSs ha a e used by he class oom eache o
communica e abou hei w i ing abili y and los ma ks. A high
sco e on hese componen s by s uden s in he good g oup
may in o m us ha hey a ibu e hei w i ing ou come o
in e nal, s able, and con ollable ac o s. On he o he hand, a
high sco e on his componen by s uden s o he poo g oup
may e eal ha hey conside hei cu en WSs as (s ill)
inadequa e o w i e a good le e (in e nal, a iable, con ollable
a ibu ion), whe eas a low sco e by membe s o he la e
g oup in o ms us ha hey may be unawa e ha hey lack hese
essen ial skills.
Quali y o he in e nal model
Table 4 ( he le hal ) p esen s he desc ip i e s a is ics o
each one o he h ee a ibu ions, and o all domain and
si ua ion-speci ic a iables, along wi h he - es s and Chi squa e
es s, sepa a ely o he ‘good’ and he ‘poo ’ g oup.
As hypo hesized, s uden s in he good g oups
endo sed signi ican ly mo e WS a ibu ions (M= 19.86;
SD = 4.3/M= 9.89; SD = 9.43; = –15.932; p<0.001) and
SRS a ibu ions (M= 10.14; SD = 2.36/M= 4.42; SD = 4.86;
= –17.64; p<0.001) han s uden s in he poo g oup.
Also, in line wi h ou expec a ions, he poo g oup endo sed
signi ican ly mo e a ec - ela ed a ibu ions (M= 5.16;
SD = 1.82/M= 7.04; SD = 1.92; = 13.938; p<0.001). These
indings sugges ha s uden s in he poo g oup a e less awa e
ha de iciency in speci ic WSs and/o in speci ic SR s a egies
may be he cause o an inadequa e w i ing pe o mance han
s uden s in he good g oup. We conclude ha hypo heses 1–3
we e con i med.
In line wi h hypo hesis 4 – s uden s in he poo g oup
sco e signi ican ly lowe han he good g oup on ou impo an
domain a iables, namely sel -e icacy, me acogni ion, mas e y
o ien a ion, and epis emic belie s. No signi ican di e ences
we e ound o ex e nal egula ion. Hypo heses 5 and 6 we e also
con i med: he sco es o he poo g oup di e ed signi ican ly
om he good g oup in he p edic ed di ec ion.
Pea son p oduc -momen co ela ions be ween he a iables
show mode a e posi i e co ela ions be ween me acogni ion
and mas e y o ien a ion goals in bo h g oups ( = 0.40/0.44)
and be ween me acogni ion and sel -e icacy in he good g oup
( = 0.44/0.23). Sel -e icacy (measu ed independen ly om he
ac ual w i ing ask) co ela es only mode a ely wi h compe ence
app aisal ( = 0.38/0.42) and weakly wi h he o he app aisals.
E o app aisal be o e doing he ask co ela es mode a ely wi h
alue app aisal ( = 0.45/0.49) and wi h e o app aisal a e
doing he ask ( = 0.48/0.47). Posi i e emo ions be o e and
a e doing he ask a e s ongly co ela ed ( = 0.72/0.75) and
so a e nega i e emo ions ( = 0.60/0.59). Only in he poo
g oup do nega i e emo ions be o e s a ing on he ask show a
mode a e in e se ela ionship wi h me acogni ion ( = –0.12/–
0.40). This is also he case o nega i e emo ions expe ienced
a e doing he ask ( = –0.10/–0.40). Nega i e emo ions
epo ed when s a ing on he w i ing ask coincided wi h
a nega i e alue app aisal ( = –0.33/–0.40), and wi h low
e o app aisal be o e ( = –0.35/–0.41) and a e he ask
( = –0.25/–0.46). No ewo hy, sa is ac ion app aisal a e doing
he ask shows a posi i e associa ion wi h posi i e emo ions
expe ienced a e he ask in bo h g oups ( = 0.33/0.38),
whe eas nega i e emo ions expe ienced a e he ask a e no
associa ed wi h sa is ac ion app aisal in he poo g oup ( = –
0.15/0.02).
In sum, he co ela ions be ween domain a iables and
app aisals and emo ions assessed in si u a e mo e p onounced
F on ie s in Educa ion 09 on ie sin.o g
educ-07-1007803 No embe 17, 2022 Time: 16:32 # 16
Boekae s e al. 10.3389/ educ.2022.1007803
addi ion, s uden s in he HIGH g oup exp essed highe posi i e
emo ions and lowe nega i e emo ions be o e and a e doing
he ask. Toge he hese esul s ell us ha s uden s who looked
back on a success ul lea ning ou come and endo sed he a ec -
ela ed a ibu ion i ems el sel -e icacious in he language
domain, had a mas e y goal o ien a ion, and expe ienced
a o able app aisals and emo ions.
ANN6 iden i ied he pe o mance indica o s as he mos
p ominen p edic o s o SRSg (see column 10). These include
he ime s uden s spen w i ing he le e (6.1), he ools hey
used (5.4), and he le el o hei echnical WSs (3.1). Apa
om hese pe o mance indica o s, ANN6 lis ed he s uden s’
sco e on mas e y goal o ien a ion (3.6), sel -e icacy (2.8),
me acogni ion (2.8) as well as e o app aisal pos (3.4) as
impo an con ibu ing p edic o s. Signi ican T- es s in o med
us u he ha he HIGH g oup sco ed signi ican ly highe
han he LOW g oup on mas e y o ien a ion, sel -e icacy, and
me acogni i e s a egies, posi i e emo ions (p e and pos ) and
on all app aisals. They sco ed signi ican ly lowe han he
con as g oup on all nega i e emo ions, excep anxie y p e and
pos . We conclude ha – simila o s uden s in he o he wo
good g oups, s uden s who endo sed he SRSg a ibu ion i ems
end o adop a mas e y goal o ien a ion, eel sel -e icacious
in he language domain, and epo ha ing access o me a-
cogni i e s a egies, howe e , hey a e uniquely cha ac e ized by
hei sco e on in es ed e o , ime spen on ask, and use o
he p o ided ools.
P edic i e con ibu ion by ca ego ies
o a iables
In he p e ious pages we mainly ocused on he p edic i e
powe o indi idual a iables ha he espec i e ANNs used
o classi y he s uden s in o he HIGH and LOW g oups.
On analyzing he mos p ominen p edic i e weigh s o he
six di e en a ibu ion g oups, we no iced speci ic pa e ns
in ol ing di e en se s o a iables. To in es iga e hese pa e ns
u he , we examined which o he six ca ego ies o a iables
ou lined in he Dual P ocessing Sel -Regula ion Model ca ied
mos p edic i e weigh s in he six ANNs (i.e. backg ound
a iables, domain a iables, con ex -sensi i e a iables, and
pe o mance indica o s). We compa ed he sou ces o he
p edic i e weigh s and hei alue o each o he a ibu ion
g oups, sepa a ely o he good and poo g oups. Resul s a e
summa ized in Table 8.
As can be seen, backg ound a iables had he g ea es impac
on A p (16.3), domain a iables had he highes pa icipa ion in
he p edic i e model o WSg (43.7), con ex -sensi i e a iables
we e ca ying he mos p edic i e weigh o he A g (40.8),
emo ions played an impo an ole in he p edic ion o A p
(26.8), especially nega i e emo ions (23.3). Finally, pe o mance
TABLE 8 Compa a i e p edic i e weigh con ibu ion o he six
a ibu ions by each o he ca ego ies o p edic o a iables.
WSp A p SRSp WSg A g SRSg Mean
Backg ound
a iables
15.1 16.3 12.8 10.3 15.0 11.7 13.6
Domain
a iables
38.3 33.7 36.7 43.7 34.5 37.7 37.7
Si ua ion
Va iables
39.3
App
12.5
E+ 3.5
E– 23.3
38.4
17.6
4.2
16.2
39.9
19.1
3.2
17.6
39.2
18.8
3.0
17.3
40.8
21.7
2.9
16.2
34.4
13.9
3.2
16.3
38.4
Pe o mance
indica o s
7.3 11.3 11.9 6.8 9.7 16.1 10.5
100% 100% 100% 100% 100% 100% 100%
App, App aisals; E+, Posi i e Emo ions; E-, Nega i e Emo ions.
indica o s we e a salien sou ce o classi y s uden s in o he
a ge g oup o SRSg (16.1).
Discussion
This a icle s a ed wi h a e-concep ualiza ion o aspec s o
Weine ’s a ibu ion heo y. We heo ized ha he a ibu ion
p ocess begins as soon as s uden s a e aced wi h a lea ning
episode and hypo hesized ha bo h he pe cep ion o he
lea ning ou come and i s concomi an causal a ibu ion come
abou h ough mul iple cogni i e and a ec i e a iables ha a e
igge ed du ing he lea ning episode. Following he guidelines
speci ied in he Dual P ocessing Sel - egula ion Model, we
p oposed ha a he beginning o a lea ning episode domain
a iables a e ac i a ed and b ough in o WM. This e okes
mul iple cogni i e app aisals and emo ions and de e mines
he QIM ha s uden s c ea e a ha momen in ime. We
in es iga ed he hypo hesis ha he quali y o he a iables
ha a e b ough in o WM du ing he lea ning p ocess
con ibu e signi ican ly o he p edic ion o oca ional s uden s’
pe cei ed lea ning ou come and i s a ibu ion. We used an
au hen ic w i ing ac i i y and eco ded he s uden s’ app aisals
and emo ions du ing he lea ning episode, including hei
pe cep ion o he lea ning ou come and hei a ibu ion o ha
ou come. We ound ha s uden s endo se any o h ee ypes o
a ibu ions, namely explaining he pe cei ed ou come in e ms
o hei own adequa e o de icien WSs, in e ms o hei use
o misuse o sel - egula ion s a egies, and / o in e ms o he
p esence o absence o a ec - ela ed a iables. Using adi ional
me hodologies as well as p edic i e sys ems me hodologies we
examined he e ec o mul iple se s o a iables on hese causal
a ibu ions. As p edic ed, mul iple eg ession analyses ga e us
some insigh in o he app aisals and emo ions ha may play
an impo an ole in he a ibu ion p ocess, bu only ANN
me hodology could de ec impo an di e ences in he pa e n
F on ie s in Educa ion 16 on ie sin.o g

educ-07-1007803 No embe 17, 2022 Time: 16:32 # 17
Boekae s e al. 10.3389/ educ.2022.1007803
o a iables ha unde lies he a ibu ion p ocess. We will i s
commen on he de ec ed pa e ns in he good g oup.
Pa e ns o domain-speci ic
in o ma ion, con ex -sensi i e
app aisals, and emo ions in he good
g oups
The QIM o s uden s in each o he good g oups is
cha ac e ized by h ee sha ed domain a iables, namely hei
sel -e icacy belie s, mas e y goal o ien a ion and access o
me acogni i e s a egies and wo sha ed app aisals, namely
a a o able compe ence app aisal and sa is ac ion app aisal.
This cha ac e iza ion implies ha s uden s in he h ee good
g oups b ing obus capaci y belie s in he language domain
and me acogni i e s a egies in o WM, as well as a mas e y goal
o ien a ion (i.e., a ocus on he de elopmen o compe ence
and imp o emen , Pin ich,2005). These cons uc i e, ac i a ed
belie s igge a o able compe ence app aisals as well as a
eeling o sa is ac ion, which Ha e (1999) e e ed o wi h he
e m ‘compe ence mo i a ion.’
Apa om hese sha ed p edic o s, each ANN lis ed
p edic o s ha uniquely de ined s uden s in ha a ibu ion
g oup. Typical o good s uden s who endo sed he WSg
a ibu ion i ems is hei epo ed di icul y app aisal, epo ed
anxie y pos , and ool use. T- es s e ealed ha s uden s who
endo sed his a ibu ion sco ed signi ican ly lowe on hese
h ee a iables han he con as g oup, bu signi ican ly highe
on in ended and epo ed e o and on posi i e emo ions p e
and pos . Gi en he na u e o he s uden s’ QIM, we iew a
WSg a ibu ion as a causal judgmen ha in o ms us ha hese
s uden s’ pe cep ion o he lea ning ou come is a ibu ed o
s able, domain-dependen belie s in he language domain and
ha cues in he ask en i onmen a e in e p e ed in acco dance
wi h hese belie s. They a e ask- ocused and p epa ed o in es
e o when he lea ning ac i i y is conside ed ‘manageable’
(i.e., when hey eel ha hey ha e he abili y o use he ules
and p ocedu es o echnical w i ing) and he cu en lea ning
ac i i y is embedded in posi i e emo ions. Un o una ely, hese
s uden s may be p one o cues ha signal an inc ease in ask
di icul y and p ocessing bu dens.
S uden s who endo sed he A g a ibu ion i ems also
b ough cons uc i e domain belie s o bea on he ask, which
in u n e oked compe ence and sa is ac ion app aisals, oge he
wi h alue and e o app aisal p e and pos , all o which
in luenced he amoun o ime spen on ask. This sugges s ha
hese s uden s a e inclined o in es ime and e o in o lea ning
ac i i ies in he language domain, p o ided ha hey uly alue
and enjoy hese ac i i ies and expe ience ew nega i e emo ions.
Un o una ely, his may imply ha hese s uden s a e sensi i e
o cues ha signal a dec ease o in insic alue o he lea ning
ask, which may e ode hei ask in ol emen .
In co espondence wi h he wo o he good g oups, he QIM
o he s uden s who endo sed he SRSg i ems unco e ed ha
hei in e p e a ion o he lea ning ou come is based on solid
domain belie s in he language domain, bu also ha hey seem
o be less a ec ed by en i onmen al cues ha signal in insic
alue and less dis u bed by cues ha signal inc eased di icul y
le el. The dis inc i e pa e n o p edic o s ha eme ged showed
he pe o mance indica o s in pole posi ion, in in e ac ion wi h
epo ed e o . These s uden s’ mind ul engagemen seems o be
cha ac e ized by hei ac i e use o sel - egula ion s a egies and
he ools p o ided by he eache . Boekae s and Co no (2005)
called his ‘e o ul accomplishmen due o oli ional con ol’
and explained ha oli ional s a egies – o good wo k habi s –
help s uden s o commi hemsel es o engage mind ully in he
lea ning ask and p o ec hei lea ning in en ion when di icul
wo k mus be comple ed. Rozendaal e al. (2003) showed
ha wo king a en i ely on lea ning asks and p o ec ing one’s
lea ning in en ion om compe ing ac ion endencies may
depend on he quali y o he implemen a ion plans ha s uden s
ha e access o and a e willing o use. Voca ional s uden s who
had been ained o implemen sel - egula ion s a egies in a
speci ic domain we e mo e mo i a ed o use hese s a egies and
epo ed less anxie y six mon hs a e he aining p og am (see
also Boekae s and Rozendaal,2006).
Pa e ns o domain-speci ic
in o ma ion, con ex -sensi i e
app aisals, and emo ions in poo
g oups
The Dual P ocessing SR model p edic s ha s uden s, who
ha e un a o able mo i a ion belie s and app aisals o a lea ning
ac i i y and expe ience concomi an nega i e emo ions mo e
on o he well-being ou e, mainly because he QIM ha hey
c ea e os e s a conce n o well-being (sel - ocus) a he han
o lea ning ( ask- ocus). Examining he QIM o he h ee poo
a ibu ion g oups shows ha hese s uden s a e no sa is ied
wi h hei lea ning ou come and may blame his on con ollable
o uncon ollable ac o s. The h ee poo g oups seem o deal
di e en ly wi h he u bulence c ea ed by he lea ning ac i i y
in hei sel - egula ion sys em.
The QIM o s uden s who epo ed a poo w i ing ou come
and endo sed he WSp a ibu ion i ems is cha ac e ized by he
na u e o hei epis emic belie s, goal o ien a ion, me acogni ion
and nega i e emo ions epo ed a he end o he lea ning
episode. They exp essed signi ican ly lowe con idence ha hey
could do he cu en ask, sco ed lowe on me acogni ion,
epo ed mo e anxie y, and admi ed ha he ask had been
ha d o do. They also epo ed highe e o in es men and ool
use han he con as g oup. Un ained obse e s may alsely
conclude ha hey had been mo e ask- ocused han he con as
F on ie s in Educa ion 17 on ie sin.o g
educ-07-1007803 No embe 17, 2022 Time: 16:32 # 18
Boekae s e al. 10.3389/ educ.2022.1007803
g oup. Howe e , as Pek un and Pe y (2014) showed, low sel -
e icacy in combina ion wi h anxie y may induce ei he eelings
o hopelessness o hope. Expe iencing loss o con ol c ea es
hopelessness and makes s uden s eso o de ensi e s a egies,
such as p oc as ina ion, ask a oidance, and disengagemen . On
he o he hand, eelings o hope will p e ail when s uden s a e
unce ain abou con ol. In he la e case hey may inc ease
e o (Goe z e al.,2010). Boekae s and Niemi i a (2005)
explained ha s uden s who epo low sel -e icacy belie s and
expe ience anxie y in e p e his as a dange signal o a h ea
o he sel and his p omp s hem o ealloca e esou ces. This
may lead o an inc ease in e o , o example by wo king as e ,
p oducing mo e ou pu , o epea ing answe s andomly in an
a emp o p e en nega i e lea ning ou comes om occu ing.
Howe e , his inc eased e o is la gely undi ec ed and should
be conside ed as an emo ion egula ion s a egy.Baumeis e e al.
(2007) con i med ha anxie y associa ed wi h a ask may shi
people’s p io i ies, o cing hem o ocus on immedia e u gen
needs, namely o escape he a e si e emo ional s a e. Ve mee
e al. (2000) demons a ed ha using dange con ol (undi ec ed
e o ) as a coping s a egy o escape he a e si e emo ional s a e
p oduced mo e a he han less e o s as well as umina ion.
This sel - ocus in e e ed wi h he ask and kep he le el o
anxie y (i.e., he dange signal) high. Using dange con ol as
an emo ion egula ion s a egy may empo a ily p o ec hese
s uden s om eache blame, because he pe cei ed nega i e
ou come is no due o lack o e o . Howe e , admi ing ha
lea ning ac i i ies in he language domain a e no one’s s eng h
and ha he cu en lea ning ac i i y was oo di icul o hem
may also ha bo some isks. I pee s in e p e i as a sign o low
abili y i may make hese s uden s ex a ulne able o bullying
and he e o e sensi i e o cues in he ask en i onmen ha
signal loss o ace. On he o he hand, a WSp a ibu ion may
p o e o be bene icial o lea ning, p o ided he s uden s uly
belie e ha he eache ea s hei a ibu ion as a o m o
eedback and uses i o c ea e p ac ice sessions ha help hem
acqui e he ules and p ocedu es o echnical w i ing ha hey
ha e no ye mas e ed.
An A p a ibu ion di e s undamen ally om a WSp
a ibu ion as he pa e n de ec ed by ANN2 clea ly shows.
As discussed in he esul sec ion, he QIM o he s uden s
who a ibu ed hei pe cei ed poo pe o mance o low A p is
domina ed by hei di icul y app aisal in combina ion wi h hei
sco es on sel -e icacy, echnical w i ing, and me acogni ion.
Dissimila o he WSp g oup, he A p g oup sco ed signi ican ly
lowe han he con as g oup on di icul y app aisal, anxie y
p e and pos , e o in es men , and ool use, despi e hei
signi ican ly lowe sco es on me acogni ion. This sugges s ha
hese s uden s a e ei he unawa e ha hey ha e ailing skills
in he language domain o ha hey we e ying o co e i
up. Inspec ion o hei QIM hin s a a delibe a e a emp o
wi hhold e o in o de o a oid loss o ace. Co ing on (2000)
labeled his emo ion egula ion s a egy ‘sel -handicapping’ and
desc ibed i as an a emp o main ain sel -wo h. Using his
s a egy helps s uden s o lowe hei le el o anxie y, p obably
because i allows hem o ex e nalize he sou ce o ailu e. I gi es
hem he illusion o con ol o e he si ua ion.
To summa ize, he di e en pa e ns be ween he a iables
ha we e unco e ed by ANN4–6 opened a window on he
s uden s’ QIM and how i is ela ed o hei pe cei ed lea ning
ou come and i s a ibu ion. I seems ha s uden s who can ely
on s able, domain-dependen mo i a ion belie s a e inclined o
conside hemsel es compe en o s a he lea ning ac i i y and
a e willing o in es e o . Cus e s and Aa s (2005) showed
ha goals and asks ha ha e been in used wi h posi i e a ec
in he pas au oma ically e oke e o di ec ed a a aining
simila goals. In o he wo ds, s uden s who engage in lea ning
ac i i ies di ec ed by hei own a o able belie s wo k a en i ely
on lea ning ac i i ies ha ha e become habi ual, because o
he posi i e a ec ha hey gene a e ( ou iniza ion o mind ul
engagemen ; Gyu ak e al.,2011).
The pa e ns be ween he p edic o s unco e ed by ANN1-
3 ell a di e gen s o y. I appea s ha s uden s in bo h poo
g oups sco e signi ican ly lowe on me acogni ion and b ing
un a o able belie s o bea on he lea ning ac i i y, bu hei
app aisals and emo ions di e undamen ally. S uden s who
a ibu ed a poo lea ning ou come o ailing WSs seemed o be
awa e (1) ha he ask was di icul , and ha ex a e o was in
o de o succeed, bu (2) ha hei e o was no good enough
o succeed, because (3) hei WSs a e (s ill) de icien . Clea y
(2011) epo ed in his espec ha such adap i e in e encing
led o imp o ed s a egic planning and be e use o p o ided
ins uc ion and ools. By con as , s uden s who a ibu e a
pe cei ed poo ou come o A p may also be awa e ha e o is
in o de o a oid losing ace, bu hey make a delibe a e a emp
o wi hhold e o , hus misleading eache and pee s – and e en
hemsel es – by e e ing o he lea ning ask as edious and no
wo h he e o .
Implica ions and limi a ions
Al hough ex ensi e esea ch has been conduc ed o
examine s uden s’ a ibu ions and ecommenda ions ha e
been o mula ed o change maladap i e causal a ibu ions
in o adap i e ones, li le is known abou he e ec o
ask demands and condi ions – and s uden s’ in e p e a ion
he eo – on hei a ibu ion o lea ning ou comes in di e en
domains. Ye , we hink ha i is heo e ically impo an
o in es iga e o wha ex en he men al ep esen a ion ha
s uden s make o a domain and o he di e en ypes o
lea ning ac i i ies wi hin ha domain a e linked o speci ic ypes
o a ibu ions.
The c i ical ques ion in ou s udy was whe he di e ences
in he QIM a e e lec ed in s uden s’ sel - epo ed a ibu ions
abou he ou come o a speci ic lea ning ac i i y. We showed
F on ie s in Educa ion 18 on ie sin.o g
educ-07-1007803 No embe 17, 2022 Time: 16:32 # 19
Boekae s e al. 10.3389/ educ.2022.1007803
in his a icle ha desc ibing s uden s’ a ibu ions exclusi ely
in e ms o Weine ’s h ee-dimensional axonomic space is no
doing jus ice o he complex message ha may be encoded
in he s uden s’ a ibu ion o a lea ning ou come. We hope
ha we ha e con inced a ibu ion esea che s ha i is
necessa y o con ex ualize a ibu ion i ems so ha hey be e
ep esen he labels ha eache s use in class o communica e
wi h s uden s abou he wha ’s, why’s and how’s o pe cei ed
lea ning ou comes. In ou opinion, he main con ibu ion o
he p esen s udy conce ns he me hod o da a collec ion and
he me hodological ools used o analyze he da a. We designed
a ques ionnai e ha apped in o he s uden s’ si ua ion-speci ic
a ibu ions. We also collec ed a wide a ay o da a, including
backg ound a iables and me acogni i e, mo i a ional, and
beha io al a iables. Apa om sel - epo s, which may no
e lec he ac ual le el o mo i a ion, we used on-line measu es
o mo i a ion and emo ion, as well as objec i ely measu ed
aspec s o pe o mance, such as ool use, ime spen , le el o
echnical WS, and luency in le e w i ing. We used adi ional
analyses, such as mul iple eg essions and he p edic i e sys ems
app oach and showed ha adi ional, linea models encoun e
di icul ies when he objec o he esea ch is o s udy he impac
o a wide ange o inpu s on he ou come a iable. In he
pas , se e al esea che s complained ha he assumed ela ion
be ween a ibu ions and mo i a ion a iables, such as mas e y
and pe o mance goals, emained sca ce and inconclusi e (see
e.g., Wol e s e al.,2013). We hink ha he eason o hese
inconclusi e indings is ha mos o he me hods o analyses
ha we e used lacked he powe o examine he e ec o
di e en ypes o a iables on a ibu ions, conside ing hem all
simul aneously. When we analyzed ou indings using mul iple
eg ession analysis, we no ed ha a g ea numbe o a iables
we e excluded om he analyses. Ou s udy demons a ed
con incingly ha ANN analysis is a mo e obus me hod o
p edic and classi y di e en ypes o a ibu ions, compa ed
wi h mo e adi ional me hodologies.
We ound ha ANN me hodology could iden i y wi h high
p ecision s uden s who used a WS,A , and SRS a ibu ion
on he basis o domain a iables, app aisals, emo ions, and
pe o mance indica o s. Due o i s o e all p ecision and he wide
scope o a iables ha we e conside ed o classi y he expec ed
a ibu ion, ANN de ec ed impo an di e ences in he ac o s
ha seem o unde lie he s uden s’ causal a ibu ions. G an ed,
ou esul s need addi ional es ing, bu wha hey show us is ha
s uden s who a ibu e a pe cei ed poo w i ing ou come o low
A p and hose who a ibu e i o poo WSp may ha e c ea ed a
di e en mind-se ha guided hei coping a emp s.
The indings o his s udy ha e impo an pedagogical
and p ac ical implica ions in seconda y educa ion. On he
one hand, i is c ucial ha eache s ecognize he di e ence
be ween a WSp and an A p a ibu ion, because – As Clea y
and Zimme man ad oca ed – he o me causal explana ion
will help s uden s o sus ain pe cep ions o e icacy in he
domain and in es e o o succeed, when p ocessing bu dens
a e ( empo a ily) inc easing. The la e asc ip ion may p omp
s uden s o e-de ine lea ning ac i i ies in ha domain, which
may change hei mo i a ion belie s and ch onically embed
simila lea ning ac i i ies in a ne wo k o nega i e emo ions
and app aisals. On he o he hand, he possibili y o iden i ying
wi h high le el o p ecision which s uden s would ha e di e en
a ibu ion pa e ns, p o ides an oppo uni y o ea ly, posi i e,
and a ge ed pedagogical in e en ions.
The e a e also a numbe o limi a ions o his s udy. Fi s
o all, he da a we e collec ed in a single measu emen wa e
and conce ned only one lea ning ac i i y. Such an app oach is
no a solid basis o uling ou al e na i e explana ions o he
obse ed ela ions. Also, he age ange (s uden s om uppe
seconda y educa ion) limi s he possibili y o gene aliza ion.
Fu he empi ical e idence is needed om s uden s o di e en
age g oups ega ding hei a ibu ions o lea ning ou comes in
he language domain.
Conclusion
A ibu ion is mo e han an indi idual di e ence a iable.
These causal judgmen s a e an in ica e pa o he SR
sys em and should be s udied in a con ex -sensi i e way.
Concep ualizing s uden s’ a ibu ions as an un olding p ocess
ha akes shape du ing a lea ning episode and is highly
in luenced by con ex ual ea u es p o ides he oppo uni y
o be e unde s and s uden s’ epis emic belie s abou e o
and abili y as he cause o (un) success ul lea ning ou comes.
Boekae s (2006) explained ha e o is a con using concep .
Fo example, low e o may be in e p e ed as a sign o li le
in e es in he lea ning ac i i y, as a signal o low abili y, as
a eac ion o pe cei ed di icul y le el o he ask, low sel -
e icacy, and anxie y. A sys em app oach has he po en ial o
b ing oge he la ge bodies o esea ch ha ha e he po en ial
o un old hese complex p ocesses. Mo eo e , i is e iden
ha changes in he s uden s’ pe cep ion o ask demands
du ing a lea ning episode may impac on he s uden s’ QIM
because cogni i e and a ec i e s a es in luence each o he
o e ime. The e o e, i is ad isable ha in u u e esea ch
on he p edic ion o a ibu ions he sequen ial and empo al
aspec s o he s uden s’ engagemen in he lea ning episode a e
aken in o accoun . This would allow esea che s o examine
whe e exac ly in he lea ning episode a ibu ions begin o
ake shape, in close connec ion wi h con ex ualized cues ha
s uden s pick up du ing he lea ning episode and he SR
s a egies hey use o coun e ac he nega i e impac o an
an icipa ed poo lea ning ou come ( o u he discussion o
hese ypes o analyses see Aze edo e al.,2011;Molenaa ,
2014;Aze edo,2015;Clea y e al.,2016;Panade o e al.,2016;
Molenaa e al.,2021). In sum, u u e esea ch needs o use he
indings om his s udy in he design o longi udinal s udies ha
F on ie s in Educa ion 19 on ie sin.o g
educ-07-1007803 No embe 17, 2022 Time: 16:32 # 20
Boekae s e al. 10.3389/ educ.2022.1007803
can cap u e he dynamic aspec s o sel - egula ion, including
un olding a ibu ion p ocesses.
Da a a ailabili y s a emen
The aw da a suppo ing he conclusions o his a icle will
be made a ailable by he au ho s, wi hou undue ese a ion.
E hics s a emen
The s udies in ol ing human pa icipan s we e e iewed
and app o ed by Resea ch E hics Commi ee o he Leiden
Uni e si y. W i en in o med consen o pa icipa e in his s udy
was p o ided by he pa icipan s’ pa en s o legal gua dian/nex
o kin.
Au ho con ibu ions
MB con ibu ed o he concep ualiza ion o he esea ch
and he design o he s udy, w o e he i s d a o he
heo e ical sec ion, and esul s and discussion. MM ca ied ou
some o he classical s a is ical analyses and some pa o he
neu al ne wo ks, w o e some pa o he in oduc ion, me hod
sec ion, ables, and con ibu ed o he manusc ip e iew. EC
con ibu ed o he concep ualiza ion o he esea ch and he
design o he s udy, ca ied ou he neu al ne wo k models,
w o e some pa o he me hodological sec ion, and e iewed
he manusc ip . All au ho s ead and app o ed he submi ed
e sion.
Funding
This s udy was unded by he Pen oe de Sel – Regula ed
W i ing P ojec (PSRWP) - Koning Willem I College om
Ne he lands. Con ac numbe : AGI-1802-2006/7 and da e:
May 15, 2007.
Con lic o in e es
The au ho s decla e ha he esea ch was conduc ed in he
absence o any comme cial o inancial ela ionships ha could
be cons ued as a po en ial con lic o in e es .
Publishe ’s no e
All claims exp essed in his a icle a e solely hose o he
au ho s and do no necessa ily ep esen hose o hei a ilia ed
o ganiza ions, o hose o he publishe , he edi o s and he
e iewe s. Any p oduc ha may be e alua ed in his a icle, o
claim ha may be made by i s manu ac u e , is no gua an eed
o endo sed by he publishe .
Supplemen a y ma e ial
The Supplemen a y Ma e ial o his a icle can be
ound online a : h ps://www. on ie sin.o g/a icles/10.3389/
educ.2022.1007803/ ull#supplemen a y-ma e ial
Re e ences
Abu Nase , S. S. (2012). P edic ing lea ne s pe o mance using a i icial neu al
ne wo ks in linea p og amming in elligen u o ing sys em. In . J. A i . In ell.
Appl. 3, 65–74. doi: 10.5121/ijaia.2012.3206
Alyahyan, E., and Dü¸s egö , D. (2020). P edic ing academic success in highe
educa ion: Li e a u e e iew and bes p ac ices. In . J. Educ. Technol. High. Educ.
17, 1–21. doi: 10.1186/s41239-020-0177-7
Ame ican Psychological Associa ion (2017). E hical p inciples o psychologis s
and code o conduc (2002, amended e ec i e June 1, 2010, and Janua y 1, 2017).
Washing on, DC: Ame ican Psychological Associa ion.
Aze edo, R. (2015). De ining and measu ing engagemen and lea ning in
science: Concep ual, heo e ical, me hodological, and analy ical issues. Educ.
Psychol. 50, 84–94. doi: 10.1080/00461520.2015.1004069
Aze edo, R., Johnson, A., Chauncey, A., and G aessen, A. (2011). “Use o
hype media o assess and con ey sel - egula ed lea ning,” in Handbook o sel -
egula ed lea ning pe o mance, eds B. Zimme man and D. Schunk (New Yo k,
NY: Rou ledge), 102–121.
Baumeis e , R. F., Zell, A. L., and Tice, D. M. (2007). “How emo ions acili a e
and impai sel - egula ion,” in Handbook o emo ion egula ion, ed. J. J. G oss
(New Yo k: The Guild o d P ess), 408–427.
Boekae s, M. (1999). Sel - egula ed lea ning: Whe e we a e oday. In . J. Educ.
Res. 31, 445–457. doi: 10.1016/S0883-0355(99)00014-2
Boekae s, M. (2002). “The on-line mo i a ion ques ionnai e: A sel - epo
ins umen o assess s uden s’ con ex sensi i i y,” in Ad ances in mo i a ion and
achie emen , Vol. 12, eds P. R. Pin ich and M. L. Maeh (New Yo k: JAI / Else ie
Science), 77–120.
Boekae s, M. (2006). “Sel - egula ion and e o in es men ,” in Handbook o
child psychology, Vol. 4, eds E. Sigel and K. A. Renninge (New Yo k: Wiley),
345–377. doi: 10.1002/9780470147658.chpsy0409
Boekae s, M. (2011). “Emo ions, emo ion egula ion, and sel - egula ion o
lea ning,” in Handbook o sel - egula ion o lea ning and pe o mance, eds B. J.
Zimme man and D. H. Schunk (London: Rou ledge/Taylo & F ancis G oup),
408–425.
Boekae s, M., and Co no, L. (2005). Sel - egula ion in he class oom: A
pe spec i e on assessmen and in e en ion. Appl. Psychol. In . Re . 54, 199–231.
doi: 10.1111/j.1464-0597.2005.00205.x
Boekae s, M., and Niemi i a, M. (2005). “Sel - egula ed lea ning: Finding a
balance be ween lea ning goals and ego-p o ec i e goals,” in Handbook o sel -
egula ion, eds M. Boekae s, P. R. Pin ich, and M. Zeidne (San Diego, CA:
Academic P ess), 417–450. doi: 10.1016/B978-012109890-2/50042-1
Boekae s, M., and Pek un, R. (2016). “Emo ion and emo ion egula ion in
academic se ings,” in Handbook o educa ional psychology, eds L. Co no and E. M.
Ande man (New Yo k: Rou ledge), 76–90.
F on ie s in Educa ion 20 on ie sin.o g
educ-07-1007803 No embe 17, 2022 Time: 16:32 # 21
Boekae s e al. 10.3389/ educ.2022.1007803
Boekae s, M., and Rozendaal, J. S. (2006). Sel - egula ion in Du ch seconda y
oca ional educa ion: Need o a mo e sys ema ic app oach o he assessmen o
sel - egula ion. Zei sch i Be u s Wi scha spädagogik Beihe 20, 49–77.
Boekae s, M., Musso, M., and Cascalla , E. (2012). Sel - egula ed lea ning and
he unde s anding o complex ou comes, ad ances in educa ional psychology and
assessmen , Vol. 2012. London: Hindawi Publishing Co po a ion, 1–3. doi: 10.
1155/2012/686385
Boekae s, M., O en, R., and Voe en, R. (2003). Examina ion pe o mance:
A e s uden ’s causal a ibu ions school-subjec speci ic? Anxie y S ess Coping 16,
331–342. doi: 10.1080/1061580031000095470
Cascalla , E., Boekae s, M., and Cos igan, T. E. (2006). Assessmen in he
e alua ion o sel - egula ion as a p ocess. Educ. Psychol. Re . 18, 297–306. doi:
10.1007/s10648-006-9023-2
Cascalla , E., Musso, M. F., Kynd , E., and Dochy, F. (2015). Modelling o
unde s anding and o p edic ion/classi ica ion – he powe o neu al ne wo ks
in esea ch. F on . Lea n. Res. 2, 67–81.
Chan, L. K. S., and Moo e, P. J. (2006). De elopmen o a ibu ional belie s
and s a egic knowledge in yea s 5-9. A longi udinal analysis. Educ. Psychol. 26,
161–185. doi: 10.1080/01443410500344209
Cizek, G. J. (2006). “S anda d se ing,” in Handbook o es de elopmen , eds
S. M. Downing and T. M. Haladyna (Mahwah, NJ: Law ence E lbaum Associa es),
225–260.
Clea y, T. J. (2011). “Eme gence o SRI mic oanalysis,” in Handbook o
sel - egula ed lea ning pe o mance, eds B. J. Zimme man and D. H. Schunk
(New Yo k, NY: Rou ledge), 329–345.
Clea y, T. J., Du ning, S. J., and A ino, A. R. (2016). Mic oanaly ic assessmen
o sel egula ed lea ning du ing clinical easoning asks: Recen de elopmen s and
nex s eps. Acad. Reason. 91, 1516–1521. doi: 10.1097/ACM.0000000000001228
Co ing on, M. V. (2000). Goal heo y, mo i a ion, and school achie emen : An
in eg a i e e iew. Ann. Re . Psychol. 51, 171–200. doi: 10.1146/annu e .psych.51.
1.171
C ombach, M. J., Boekae s, M., and Voe en, M. J. M. (2003). Online
measu emen o app aisals o s uden s aced wi h cu icula asks. Educ. Psychol.
Measu e. 63, 96–111. doi: 10.1177/0013164402239319
Cus e s, R., and Aa s, H. (2005). Posi i e a ec as implici mo i a o : On he
nonconscious ope a ion o beha io al goals. J. Pe s. Soc. Psychol. 89:129. doi:
10.1037/0022-3514.89.2.129
De ienne, K. B., De ienne, D. H., and Joshi, S. A. (2003). Neu al ne wo ks as
s a is ical ools o business esea che s. O gan. Res. Me hods 6, 236–265. doi:
10.1177/1094428103251907
Dowson, M., and McIne ney, D. M. (2004). The Goal o ien a ion and
lea ning s a egies su ey. Educ. Psychol. Measu e. 64, 290–310. doi: 10.1177/
0013164403251335
E klides, A. (2011). In e ac ions o me acogni ion wi h mo i a ion and a ec in
sel - egula ed lea ning: The MASRL model. Educ. Psychol. 46, 6–25. doi: 10.1080/
00461520.2011.538645
Fong, S., Si, Y.-W., and Biuk-Aghai, R. P. (2009). “Applying a hyb id model o
neu al ne wo k and decision ee classi ie o p edic ing uni e si y admission,” in
P oceedings o he 7 h In e na ional Con e ence on In o ma ion, Communica ion,
and Signal P ocessing (ICICS2009) (Macau: IEEE P ess), 1–5. doi: 10.1109/ICICS.
2009.5397665
Goe z, T., F enzel, A. C., S oege , H., and Hall, N. C. (2010). An eceden s o
e e yday posi i e emo ions: An expe ience sampling analysis. Mo i . Emo . 34,
49–62. doi: 10.1007/s11031-009-9152-2
G aham, S. (2020). An a ibu ion heo y o mo i a ion. Con emp. Educ. Psychol.
61, 1–11. doi: 10.1016/j.cedpsych.2020.101861
G aham, S., and Taylo , A. Z. (2014). “An a ibu ional app oach o
emo ional li e in he class oom,” in Handbook o emo ions in educa ion,
eds R. Pek un and L. Linnenb ink-Ga cia (New Yo k: Taylo & F ancis),
96–119.
Gyu ak, A., G oss, J. J., and E kin, A. (2011). Explici and implici emo ion
egula ion: A dual-p ocess amewo k. Cogn. Emo . 25, 400–412. doi: 10.1080/
02699931.2010.544160
Ha e , S. (1999). The cons uc ion o he sel . New Yo k: Guil o d P ess.
Ho e , B., and Pin ich, P. R. (2002). Pe sonal epis emology: The pyschology o
belie s abou knowledge and knowing. New Yo k, NY: Rou ledge. doi: 10.4324/
9781410604316
Kanakana, G., and Olan ewaju, A. (2011). “P edic ing s uden pe o mance in
enginee ing educa ion using an a i icial neu al ne wo k a Tshwane Uni e si y o
Technology,” in P oceedings o he ISEM (S ellenbosch).
Kel ne , D., Oa ley, K., and Jenkins, J. M. (2013). Unde s anding emo ions.
New Yo k: John Wiley.
Lippmann, R. P. (1987). An in oduc ion o compu ing wi h neu al ne s. IEEE
ASSP Mag. 3, 4–22.
Molenaa , I. (2014). Ad ances in empo al analysis in lea ning and ins uc ion.
F on . Lea n. Res. 2, 15–24. doi: 10.14786/ l . 2i4.118
Molenaa , I., Ho e s, A., and Bake , R. S. (2021). Wha can momen -by-
momen lea ning cu es ell us abou sel - egula ed lea ning? Lea n. Ins .
72101206. doi: 10.1016/j.lea nins uc.2019.05.003
Musso, M. F., He nández, C. F. R., and Cascalla , E. C. (2020). P edic ing key
educa ional ou comes in academic ajec o ies: A machine-lea ning app oach.
High. Educ. 80, 875–894. doi: 10.1007/s10734-020-00520-7
Musso, M. F., Kynd , E., Cascalla , E. C., and Dochy, F. (2013). P edic ing
gene al academic pe o mance and iden i ying he di e en ial con ibu ion o
pa icipa ing a iables using a i icial neu al ne wo ks. F on . Lea n. Res. 1, 42–71.
doi: 10.14786/ l . 1i1.13
Musso, M., Kynd , E., Cascalla , E., and Dochy, F. (2012). P edic ing
ma hema ical pe o mance: The e ec o cogni i e p ocesses and sel - egula ion
ac o s. Educ. Res. In . 2012, 1–13. doi: 10.1155/2012/250719
Neal, W., and Wu s , J. (2001). Ad ances in ma ke segmen a ion. Ma k. Res.
13, 14–18.
Panade o, E., Klug, J., and Ja ela, S. (2016). Thi d wa e o measu emen
in he sel egula ed lea ning ield: When measu emen and in e en ion come
hand in hand. Scand. J. Educ. Res. 60, 723–735. doi: 10.1080/00313831.2015.106
6436
Pek un, R., and Pe y, R. P. (2014). “Con ol- alue heo y o achie emen
emo ions,” in Handbook o emo ions in educa ion, eds R. Pek un and L.
Linnenb ink-Ga cia (New Yo k: Taylo & F ancis), 120–141. doi: 10.4324/
9780203148211
Pin ich, P. R. (2000). “The ole o goal-o ien a ion in sel - egula ed lea ning,” in
Handbook o sel - egula ion, eds M. Boekae s, P. R. Pin ich, and M. Zeidne (San
Diego, CA: Academic P ess), 451–502. doi: 10.1016/b978-012109890-2/50043-3
Pin ich, P. R. (2005). “The ole o goal o ien a ion in sel - egula ed lea ning,”
in Handbook o Sel - egula ion, eds M. Boekae s, P. R. Pin ich, and M. Zeidne
(San Diego, C.A: Academic P ess), 451–502. doi: 10.1016/B978-012109890-2/50
043-3
Pin ich, P. R., and De G oo , E. V. (1990). Mo i a ional and sel - egula ed
lea ning componen s in class oom academic pe o mance. J. Educ. Psychol. 82,
33–40. doi: 10.1037/0022-0663.82.1.33
Rod iguez-He nandez, C. F., Musso, M. F., Kynd , E., and Cascalla , E. C.
(2021). A i icial neu al ne wo ks in academic pe o mance p edic ion: Sys ema ic
implemen a ion and p edic o e alua ion. Compu . Educ. A i . In ell. 2:100018.
doi: 10.1016/j.caeai.2021.100018
Rozendaal, J. S., Minnae , A., and Boekae s, M. (2003). Mo i a ion and sel -
egula ed lea ning in seconda y oca ional educa ion: In o ma ion-p ocessing
ype and gende di e ences. Lea n. Indi id. Di . 13, 273–289.
Schomme , M. (1990). E ec s o belie s abou he na u e o knowledge on
comp ehension. J. Educ. Psychol. 82, 498–504. doi: 10.1037/0022-0663.82.3.498
Schunk, D. H., and Zimme man, B. J. (Eds). (1998). Sel - egula ed lea ning:
F om eaching o sel - e lec i e p ac ice. New Yo k, NY: Guil o d P ess.
Seege s, G., and Boekae s, M. (1993). Task mo i a ion and ma hema ics
achie emen in ac ual ask si ua ions. Lea n. Ins . 3, 133–150. doi: 10.1016/0959-
4752(93)90012-O
Ve mee , H., Boekae s, M., and Seege s, G. (2000). Mo i a ional and gende
di e ences: Six h-g ade s uden s’ ma hema ical p oblem-sol ing beha io . J. Educ.
Psychol. 92, 308–315. doi: 10.1037/0022-0663.92.2.308
Ve mun , J. (1998). The egula ion o cons uc i e lea ning p ocesses.
B . J. Educ. Psychol. 68, 49–171. doi: 10.1111/j.2044-8279.1998. b01
281.x
Vole , S. (2001). “Unde s anding lea ning and mo i a ion in con ex : A
mul i-dimensional and mul i-le el cogni i e-si ua i e pe spec i e,” in Ad ances
in lea ning and ins uc ion se ies: Mo i a ion in lea ning con ex s, eds S.
Vole and S. Jä elä (Ams e dam: Eme ald G oup Publishing Limi ed),
57–82.
Weine , B. (2018). The legacy o an a ibu ion app oach o mo i a ion and
emo ion: A no-c isis zone. Mo i . Sci. 4, 4–14. doi: 10.1037/mo 0000082
Whi e, H., and Racine, J. (2001). S a is ical in e ence, he boo s ap, and neu al
ne wo k modelling wi h applica ion o o eign exchange a es. IEEE T ans. Neu al
Ne w. 12, 657–673. doi: 10.1109/72.935080
F on ie s in Educa ion 21 on ie sin.o g

educ-07-1007803 No embe 17, 2022 Time: 16:32 # 22
Boekae s e al. 10.3389/ educ.2022.1007803
Wol e s, A., Fan, W., and Daughe y, S. G. (2013). Examining
achie emen goals and causal a ibu ions oge he as p edic o s o academic
unc ioning. J. Exp. Educ. 81, 295–321. doi: 10.1080/00220973.2012.
700498
Zimme man, B., and Clea y, T. J. (2009). “Mo i es o sel - egula ed lea ning:
A social cogni i e accoun ,” in Mo i a ion a school, eds K. R. Wen zel and A.
Wig ield (NewYo k, NY: Rou ledge), 11–34.
Zimme man, B., and Ki san as, A. (2002). Acqui ing w i ing e ision and sel -
egula o y skill h ough obse a ion and emula ion. J. Educ. Psychol. 94, 660–668.
doi: 10.1037/0022-0663.94.4.660
Zimme man, B., and Schunk, D. H. (2011). “Sel - egula ed lea ning and
pe o mance: An in oduc ion and an o e iew,” in Handbook o sel - egula ion
o lea ning and pe o mance, 1s Edn, eds D. H. Schunk and B. Zimme man
(NewYo k, NY: Rou ledge), 1–12. doi: 10.4324/9780203839010
F on ie s in Educa ion 22 on ie sin.o g