HeLP: The Hebrew Lexicon project

Vol.:(0123456789)
Beha io Resea ch Me hods (2024) 56:8761–8783
h ps://doi.o g/10.3758/s13428-024-02502-4
ORIGINAL MANUSCRIPT
HeLP: The Heb ew Lexicon p ojec
RoniS ein1· RamF os 1,2· NoamSiegelman1
Accep ed: 17 July 2024 / Published online: 9 Sep embe 2024
© The Au ho (s) 2024
Abs ac
Lexicon p ojec s (LPs) a e la ge-scale da a esou ces in di e en languages ha p esen beha io al esul s om isual wo d
ecogni ion asks. Analyses using LP da a in mul iple languages p o ide e idence ega ding c oss-linguis ic di e ences as
well as simila i ies in isual wo d ecogni ion. He e we p esen he i s LP in a Semi ic language— he Heb ew Lexicon
P ojec (HeLP). HeLP assembled lexical decision (LD) esponses o 10,000 Heb ew wo ds and nonwo ds, and naming
esponses o a subse o 5000 Heb ew wo ds. We used he la ge-scale HeLP da a o es ima e he impac o gene al p edic o s
(lexicali y, equency, wo d leng h, o hog aphic neighbo hood densi y), and Heb ew-speci ic p edic o s (Semi ic s uc u e,
p esence o cli ics, phonological en opy) o isual wo d ecogni ion pe o mance. Ou esul s e ealed he ypical e ec s o
lexicali y and equency ob ained in many languages, bu mo e complex impac o wo d leng h and neighbo hood densi y.
Conside ing Heb ew-speci ic cha ac e is ics, HeLP da a e ealed be e ecogni ion o wo ds wi h a Semi ic s uc u e han
wo ds ha do no con o m o i , and a d op in pe o mance o wo ds comp ising cli ics. These e ec s a ied, howe e , ac oss
LD and naming asks. Las ly, a signi ican inhibi o y e ec o phonological ambigui y was ound in bo h naming and LD.
The implica ions o hese indings o unde s anding eading in a Semi ic language a e discussed.
Keywo d Visual wo d ecogni ion; Mega s udies; Reading; C oss-linguis ic di e ences
In oduc ion
The abili y o apidly iden i y p in ed sequences o le e s as
indi idual wo ds and au oma ically access hei phonologi-
cal and seman ic ep esen a ions has in igued scien is s o
decades, and is s ill he ocus o ex ensi e esea ch in cogni-
i e science. Tasks ha measu e pa icipan s eac ion imes
(RTs) and ecogni ion accu acy o p in ed le e sequences
ha e p o ided impo an insigh s ega ding he compu a ions
unde lying isual wo d ecogni ion. Pe haps he mos common
expe imen al pa adigm in such s udies is he lexical decision
(LD) ask, whe e pa icipan s a e p esen ed wi h le e s ings,
one a a ime, and a e equi ed o p o ide as esponses as o
whe he o no hey ep esen exis ing wo ds. Ano he common
ask is he naming ask, in which pa icipan s a e equi ed o
p onounce, as as and as accu a ely as possible, a isually
p esen ed wo d. RT and accu acy da a using he LD and nam-
ing asks we e aken o e eal he unde lying compu a ions
in ecognizing p in ed wo ds p esen ed in isola ion, e ie -
ing hei phonological s uc u e, and accessing hei seman ic
ep esen a ion. Ac oss many s udies, hey demons a ed highly
eplicable e ec s which oday a e he landma ks o isual wo d
ecogni ion: As a non-exhaus i e lis , wo ds a e ecognized
as e han nonwo ds (i.e., he lexicali y e ec , e.g., Fo s e &
Chambe s, 1973; Monsell e al., 1989); equen wo ds a e ec-
ognized as e han in equen wo ds (i.e., he wo d equency
e ec , e.g., B oadben , 1967; B ysbae e al., 2017); sho e
wo ds a e p ocessed as e han longe wo ds (i.e., he wo d
leng h e ec , e.g., F ed iksen & K oll, 1976; Hudson & Be g-
man, 1985); and he numbe and equency o o hog aphic
neighbo s a ec decision imes (i.e., he neighbo hood densi y
e ec ; e.g., And ews, 1992; G ainge e al., 1989).
Lexicon p ojec s indi e en languages
In he i s decades o isual wo d ecogni ion s udies,
esea che s ypically employed expe imen al designs in
which e bal s imuli we e selec ed o ep esen ac o s o
* Noam Siegelman
[email p o ec ed]
1 Depa men o Psychology, The Heb ew Uni e si y
o Je usalem, Moun Scopus Campus, 9190501Je usalem,
Is ael
2 BCBL, Basque Cen e o Cogni ion, B ain andLanguage,
SanSebas ian, Spain
8762 Beha io Resea ch Me hods (2024) 56:8761–8783
in e es (e.g., imageabili y, conc e eness, mo phological
complexi y, homog aphy, homophony), measu ing pe -
o mance o hese s imuli. Gene ally, hese s udies used
a ela i ely small se o s imuli, o en no ep esen a i e
o he a ie y ound ac oss he ull lexicon, hus lim-
i ing hei ex e nal alidi y. One p ominen example is
he ocus on monosyllabic wo ds in English in isual
wo d ecogni ion s udies, despi e he ac ha in many
languages hey ep esen a small po ion o he wo ds
in he language (e.g., less han 15% o wo ds, Fe and
e al., 2010).
An elegan solu ion o hese limi a ions was he Eng-
lish Lexicon P ojec (ELP, Balo a e al., 2007), which
p esen ed an open la ge-scale da a esou ce ha included
o e 40,000 wo ds along wi h hei espec i e beha io al
da a in bo h he LD and he naming asks. The ELP made
i possible, o he i s ime, o es a ange o hypo h-
eses ega ding wo d ecogni ion compu a ions wi hou he
need o cons uc a a ge ed expe imen wi h i s ine i able
limi a ions. Ins ead, esea che s could gene a e hypo heses
ega ding he impac o any a iable, and simply ex ac
he beha io al da a o all ele an s imuli om he da a-
base (a p ocess some imes e e ed o as “ i ual expe i-
men s”, see, e.g., Kupe man, 2015). The ELP has been
used ex ensi ely since i s publica ion in o de o explo e
he unde lying compu a ions o wo d ecogni ion in Eng-
lish, and o da e has been ci ed o e 3000 imes. I has been
use ul in alida ing and es ing he impac o a ange o
psycholinguis ic ac o s on eading, such as wo ds’ seman-
ic anspa ency (Kim e al., 2018), wo d leng h (e.g., New
e al., 2006), imageabili y (e.g., Dyma ska e al., 2023);
o hog aphic–phonological egula i ies (e.g., Chee e al.,
2020; Siegelman e al., 2020), and o hog aphic–seman ic
consis ency (Siegelman e al., 2022).
Impo an ly, he ELP has also inspi ed he c ea ion
o pa allel lexicon p ojec s (LPs) in o he languages, o
p o ide a c i ical c oss-linguis ic pe spec i e in eading
esea ch. These LPs include B i ish English, o be dis in-
guished om Ame ican English (Keulee s e al., 2012);
Du ch (Keulee s e al., 2010); F ench (Fe and e al.,
2010); Spanish (Aguas i as e al., 2018); Pe sian (Nema i
e al., 2022); Malay (Yap e al., 2010); Ge man (Sch eu e
& Sch oede , 2017); Po uguese (Soa es e al., 2019); and
Chinese (Tse e al., 2017). Toge he , he wide scale o
c oss-linguis ic da a o e ed by he di e en LPs ha e p o-
ided impo an insigh s in o how basic wo d ecogni ion
p ocesses a y ac oss languages, and how he p ope ies o
a gi en w i ing sys em shape he cogni i e compu a ions
in ol ed in eading one language compa ed o ano he .
F om his pe spec i e, he speci ic cha ac e is ics o a
w i ing sys em can be aken as an “expe imen al manipu-
la ion” o examine hei impac on eading pe o mance
(see F os , 2012, o discussion).
Bu no e ha whe eas all LPs ha e he same aim, me h-
odologically he e is subs an ial a iabili y in how hey we e
cons uc ed. Fo example, while all a ailable LPs p o ide
LD da a, no all include naming da a. Fu he , al hough all
LPs do p o ide a b oade co e age o a language’s wo ds
han ypical expe imen s, he e is s ill subs an ial a i-
abili y in he numbe o wo ds hey employed (e.g., om
1800 wo ds and nonwo ds in Pe sian, Nema i e al., 2022;
o 40,481 wo ds and nonwo ds in English, Balo a e al.,
2007). LPs also di e in he numbe o pa icipan s p o-
iding esponses o each wo d in he p ojec (e.g., om 25
esponses pe wo d in he F ench LP, Fe and e al., 2010, o
300 in he Spanish LP, Aguas i as e al., 2018), and in o he
design cha ac e is ics such as he wo d–nonwo d a io in
he LD ask. Table1 summa izes he me hodological p op-
e ies o exis ing LPs, as well as he cha ac e is ics o he
languages s udied.
In spi e o he subs an ial me hodological a iabili y,
la ge-scale LP da a ha e none heless la gely eplica ed
mul iple well-es ablished isual wo d ecogni ion e ec s
ac oss languages, demons a ing subs an ial simila i ies in
compu a ions. Fo example, LPs ha e consis en ly e ealed
equency and lexicali y e ec s, as well as o hog aphic
neighbo hood densi y e ec s. Impo an ly, howe e , he
mul ilingual compa isons ha LPs inspi ed ha e also dem-
ons a ed signi ican di e ences in eading beha io ac oss
languages, p o iding illumina ing insigh s in o how he
p ope ies o a language and i s w i ing sys em shape he
compu a ions in ol ed in p ocessing p in ed in o ma ion.
Fo example, LD da a om he Pe sian LP showed no wo d
leng h e ec , in con as o o he LPs. This has led au ho s
o assume ha he speci ic p ope ies o he Pe sian o hog-
aphy—speci ically, he unde -speci ica ion o owels in
he p in ed language o he s ong co ela ion be ween wo d
leng h and o hog aphic neighbo hood size—a e he ea-
son o he absence o he wo d leng h e ec (Nema i e al.,
2022). In s a k con as o Pe sian, he Malay LP e ealed
ha he leng h o wo ds in Malay is he s onges p edic o
o wo d ecogni ion ime. This again was ied o he s uc u e
o he w i ing sys em: he Malay language has a shallow
o hog aphy, anspa en mo phology, and a simple syllabic
s uc u e, which p esumably lead Malay eade s o ely p i-
ma ily on se ial con e sion o le e s o phonemes du ing
wo d ecogni ion. Indeed, u he analysis o he Malay LP
e ealed g ea di e ences be ween Malay and English in he
con ibu ion o p edic o s such as equency, o hog aphic
neighbo s, and wo d leng h o wo d ecogni ion pe o mance
(Yap e al., 2010).
The c oss-linguis ic a ia ions e ealed by la ge-scale
da a highligh how he basic p ope ies o a w i ing sys em
impac wo d ecogni ion beha io ac oss languages, leading
o a deep unde s anding o he uni e sal p inciples in ol ed
in o hog aphic p ocessing ac oss he wo ld’s languages.
8763Beha io Resea ch Me hods (2024) 56:8761–8783
Table 1 A summa y o la ge-scale Lexicon P ojec in di e en languages, and he p ope ies o hose languages
Linguis ic cha ac e is ics Lexicon p ojec de ails
Lexicon P ojec Topological amily
(b anch)
Sc ip (sc ip ype) Mo phological ypol-
ogy
O hog aphic
anspa ency
Numbe o wo ds LD/naming ask A e age obse a ions
pe wo d
Commen s
Chinese, Manda in
(simpli ied sc ip )
(Tsang e al., 2018)
Sino-Tibe an (Sini ic) Chinese (logo-
g aphic)
Analy ic Opaque 12,578 LD 42 See also Sze e al.,
2014
Chinese, Can onese
( adi ional sc ip )
(Tse e al., 2017)
Sino-Tibe an (Sini ic) Chinese (logo-
g aphic)
Analy ic Opaque 25,286 LD 33 All s imuli we e wo-
cha ac e compound
wo ds
Du ch (Keulee s
e al., 2010)
Indo-Eu opean (Wes
Ge manic)
La in (alphabe ic) Syn he ic, usional Mode a e 14,089 LD 39 Monosyllabic and
disyllabic wo ds
English (Ame ican)
(Balo a e al., 2007)
Indo-Eu opean (Wes
Ge manic)
La in (alphabe ic) Mode a ely analy ic Opaque 40,481 LD & naming LD- 34, Naming- 25
English (B i ish)
(Keulee s e al.,
2012)
Indo-Eu opean (Wes
Ge manic)
La in (alphabe ic) Mode a ely analy ic Opaque 14,365 LD 39 Monosyllabic and
disyllabic wo ds
F ench (Fe and
e al., 2010)
Indo-Eu opean
(Romance)
La in (alphabe ic) Mode a ely analy ic Mode a e 38,840 LD 25
Ge man (Sch ö e &
Sch oede , 2017)
Indo-Eu opean (Wes
Ge manic)
La in (alphabe ic) Syn he ic, usional Mode a e 1152 LD & naming LD- 18, Naming- 20 De elopmen al p ojec :
Pa icipan s om 1s
g ade s o adul s
Malay (Yap e al.,
2010)
Aus onesian
(Malayo-Polyne-
sian)
La in (alphabe ic) Syn he ic, agglu ina-
i e
T anspa en 9592 LD & naming 44 Pa icipan s go
eedback i hey we e
w ong
Pe sian (Nema i
e al., 2022)
Indo-Eu opean
(Wes e n I anian)
A abic Syn he ic, usional Abjad 1800 LD Wo ds 3–8 le e s long
Eu opean Po u-
guese (Soa es e al.,
2019)
Indo-Eu opean
(Romance)
La in (alphabe ic) Syn he ic, usional Opaque 1920 LD & naming 55
Spanish (Aguas i as
e al., 2018)
Indo-Eu opean
(Romance)
La in (alphabe ic) Syn he ic, usional T anspa en 44,853 LD 333 Wo d o nonwo ds
a io o 7:3
8764 Beha io Resea ch Me hods (2024) 56:8761–8783
Since languages na u ally di e in hei sc ip s and how
hei w i ing sys ems ep esen sound and meaning, a good
heo y o eading should be able o explica e how his a ia-
ion impac s he p ocessing o p in ed wo ds. When a gi en
language di e ges om o he s in e ms o eade s’ pe o -
mance, impo an e idence is u nished ega ding uni e sal
p inciples o eading (see F os , 2012, o discussion). Such
esea ch en e p ise, howe e , equi es la ge-scale da a om
many di e se w i ing sys ems. The main goal o he p e-
sen mega-s udy is o con ibu e o his impo an esea ch
e o by p o iding, o he i s ime, sys ema ic da a om a
Semi ic language—Heb ew—a language ha has o en been
shown o p oduce con as i e esul s ela i e o Eu opean
languages.
The Heb ew language: Animpo an es case
Heb ew is a Semi ic language, as a e A abic, Amha ic, and
Mal ese. Many wo ds in Semi ic languages a e oo -de i ed,
so ha hei base is a oo mo pheme, usually consis ing o
h ee consonan s, which con eys he co e meaning o he
wo d. Semi ic wo ds a e cons uc ed by in e wining oo
mo phemes wi h wo d pa e n mo phemes—abs ac phono-
logical s uc u es, consis ing o owels o o owels and con-
sonan s, in which he e a e “open slo s” o he oo ’s con-
sonan s o i in o. In gene al, wo d pa e ns p o ide ague
mo phosyn ac ic in o ma ion. Fo example, in Heb ew, he
oo K.Š.R., which con eys he gene al no ion o “ ying”,
and he wo d pa e n / i–o-e /, which is mos ly used o deno e
eminine nouns, o m he wo d / ikšo e /, meaning “com-
munica ion”. Embedding he oo K.Š.R. in he wo d pa e n
/-i-u-/ p oduces he wo d /kišu /, meaning “link”, e c. The
oo consonan s can be dispe sed wi hin he wo d in many
possible posi ions, and he e is li le a p io i in o ma ion
ega ding hei loca ion. Wo d pa e ns ha e a well-de ined
in e nal s uc u e. Thei onse comp ises a es ic ed numbe
o consonan s (mainly /h/, /m/, / /, /n/, /l/), and he o de
and iden i y o subsequen consonan s and owels is igid.
Because he e a e no a p io i cons ain s ega ding he loca-
ion o oo consonan s in he wo d, he main clue ega ding
hei iden i y is he well-de ined phonological s uc u e o
he wo d pa e n ha allows he oo consonan s o s and
ou (Deu sch e al., 1998, 2021; Lado -Weizman & Deu sch,
2022). O e all, he e a e abou 3000 oo s in Heb ew, abou
100 nominal wo d pa e ns, and se en e bal pa e ns.
Ano he majo cha ac e is ic o he Heb ew w i ing sys-
em is i s ex eme phonological unde -speci ica ion. Heb ew
p in consis s o 22 le e s, o which i e ha e a ini e le e
o m. The le e s ep esen mos ly consonan al in o ma ion,
and mos owel in o ma ion is no con eyed in p in (see
Shim on, 2006; Ra id, 2011, o e iew). Two le e s—one
o bo h /o/ and /u/, and one o /i/—may con ey he owel
in o ma ion; howe e , in ce ain con ex s, hese le e s also
con ey he consonan s / / and /y/, espec i ely. This esul s
in hea y phonological decoding demands, since a subs an ial
pa o he phonological in o ma ion is missing. The missing
owels lead o an ex ensi e homog aphy, wi h many p in ed
le e s ings ha ing mul iple p onuncia ions and meanings
(e.g., he p in ed wo d “ספר”, “SFR”, mos commonly ead
as /se e /, has se en possible p onuncia ions, each wi h a
di e en meaning, depending on di e en owel con igu a-
ions). Bu since he s uc u e o spoken wo ds is highly
cons ained by he ela i ely small numbe o Semi ic wo d
pa e ns, eade s can con e ge on a gi en wo d qui e easily
du ing ex eading, since he p in ed o m ypically de e -
mines he app op ia e wo d pa e n wi h ela i ely high eli-
abili y, and once a wo d pa e n has been ecognized, he ull
owel in o ma ion is a ailable o he eade , e en i i is no
speci ied by he p in ed o m (Deu sch e al., 2021; F os ,
2006). Hence, Heb ew p in p o ides a pe ec example o
op imiza ion o in o ma ion, whe e subs an ial mo phologi-
cal (and he e o e seman ic) in o ma ion is p o ided along
wi h su icien phonological cues using minimal o ho-
g aphic symbols (see F os , 2012).1
F om he pe spec i e o o hog aphic dep h (F os e al.,
1987; Ka z & os , 1992; Schmalz e al., 2015), Heb ew is
conside ed o ha e a e y deep o hog aphy, gi en mul iple
ea u es ela ed o an unde ep esen a ion o phonology in
p in . The i s is he missing owel in o ma ion discussed
abo e. In con as o he dep h o he English w i ing sys em,
which esul s mainly om phonological inconsis ency o
owel le e s (e.g., EA is p onounced di e en ly in DEAR,
HEAD, and STEAK), in Heb ew, he owel in o ma ion is
gene ally no inconsis en bu missing. This poses a challenge
in measu ing and de ining he phonological unce ain y o
Heb ew wo ds (F os , 1994, 1995). A second sou ce o pho-
nological unce ain y in he Heb ew o hog aphy is a eed-
o wa d and eed-backwa d inconsis ency o a ew consonan al
le e s, whe e some can be mapped o di e en phonemes, and
some phonemes can be ep esen ed by di e en le e s. Fo
example, he Heb ew le e “כ” can be p onounced as /x/ o
/k/, ״ב״ can be p onounced as / / o /b/, and ‘פ’ can be p o-
nounced as / /, o /p/. Con e sely, / / can be w i en wi h he
le e s "ת"o "ט", /s/ can be w i en as “ס” and “ש”, /k/ can be
w i en as ‘כ’’ and “ק”, and /x/ can be w i en as “כ” and “ח”.
Las ly, ano he impo an ea u e o Heb ew is ha mo -
pho-syn ac ic in o ma ion ha is con eyed in Eu opean
1 Heb ew also has a diac i ical a ian o he w i ing sys em, whe e
owels a e depic ed by poin s (appea ing mos ly unde p in ed le -
e s). These owel ma ks a e ypically augh in he i s g ade, assis -
ing eache s in de eloping decoding skills du ing eading acquisi ion,
bu s a ing om he end o he second g ade, p in ed and w i en
Heb ew does no no mally include diac i ical ma ks (see Sha e &
Ba -On, 2018, o a de ailed discussion). Apa om poe y and eli-
gious ex s, adul eading ma e ial is always un-poin ed.
8765Beha io Resea ch Me hods (2024) 56:8761–8783
languages by unc ion wo ds (e.g., “ he”, “ om”, “ o”,
“and”) is con eyed in Heb ew as single le e s which a e
a ached o he wo d (i.e., cli ics; e.g., “ he” = > ה, “ om
“ = > מ, “ o “ = > ל, “and” = > ו). Fo example, he ou -
wo d English sequence “and om he house” is p in ed in
Heb ew as one wo d, whe e he h ee le e s con eying and/
om/ he a e a ached o he base wo d “house” (“״ומהבית,
p in ed as VMHBYT, ead as / emehabayi /, see Ra id,
2011). Cli ics a e abundan in Heb ew p in ed inpu ,
al hough hei impac on isual wo d ecogni ion is cu en ly
la gely unknown. This is due o he endency o p e ious
s udies in he wo d ecogni ion li e a u e (in Heb ew as in
o he languages) o ocus on he p ocessing o base wo ds,
a he han on mo e na u alis ic s imuli which be e e lec
he dis ibu ion o p in ed wo ds in a language.
Visual wo d ecogni ion inHeb ew: Ase ies
o di e gen indings
Gi en he unique cha ac e is ics o he Heb ew w i ing
sys em, i has been he ocus o ex ensi e esea ch in es i-
ga ing how hese impac eading. In ac , Heb ew has been
used as a case s udy o in es iga ing a a ie y o ele an
domains, such as he p edic o s o eye mo emen s du -
ing eading (e.g., Deu sch e al., 2003; Velan e al., 2013),
he s udy o he de elopmen al ajec o y o li e acy (e.g.,
Sha e, 1999; Sha e & Ba -on, 2018), he impac o di e en
sc ip s on eading in a second language (e.g., Mo & P io ,
2020, 2021), and how he o hog aphic s uc u e o Heb ew
is e lec ed in a ious o ms o dyslexia (e.g., F iedmann
& Luko , 2008; F iedmann & Rahamim, 2007). Gi en he
ocus on he cu en s udy, howe e , ou e iew cen e s on
how he p ope ies o he Heb ew w i ing sys em impac
isual wo d ecogni ion p ocesses. Gene ally, and in line
wi h ea lie s udies in English and o he languages desc ibed
abo e, p e ious wo k has ended o ocus on how speci ic
ea u es o he Heb ew o hog aphy (e.g., homog aphy, pho-
nological ambigui y, mo phological s uc u e) lead o di e -
gen pa e ns o isual wo d p ocessing. Then, he impac
(o lack he eo ) o he s udied ea u e was discussed wi hin
a b oade amewo k, owa ds unde s anding he uni e sal
p inciples ha d i e wo d ecogni ion p ocesses ac oss lan-
guages. Unde lying his esea ch agenda is he heo e ical
supposi ion ha i language A (e.g., English) shows a gi en
pa e n o beha io , and language B (e.g., Heb ew) does no ,
his poin s o a highe -o de p inciple ha simul aneously
explains bo h phenomena.
Wi hin his esea ch en e p ise, subs an ial wo k has
ocused on Heb ew’s ex eme phonological unde -speci i-
ca ion and homog aphy. Some s udies sugges ed ha lexical
decisions in Heb ew a e made p io o phonological dis-
ambigua ion (Ben in & F os , 1987), and e lec he com-
pu a ion o a phonological impo e ished code (see F os ,
1998, o discussion). Using he naming ask, F os e al.
(1987) showed ha equency and seman ic p iming e ec s
a e s onge in Heb ew han in languages wi h shallowe
o hog aphies, esembling he e ec s e ealed in LD. These
esul s we e aken o indica e ha eade s o languages wi h
deep o hog aphies such as Heb ew ely mo e hea ily on
lexical in o ma ion han eade s o languages wi h shallow
o hog aphies (e.g., Finnish, Spanish, Ge man, Du ch), in
o de o compu e phonology om p in . This ea ly ind-
ing is in line wi h he claim ha eade s o di e en w i -
ing sys ems ely on di e en in o ma ional cues in ligh o
hei w i ing sys em’s s uc u e (e.g., Hi sho n & Ha is,
2022; Lallie & Ca ei as, 2018; Rau e al., 2015; Seiden-
be g, 2011; Seymou e al., 2003). Howe e , all isual wo d
ecogni ion s udies in Heb ew in ol ed only a ew dozen
wo ds ha we e selec ed as s imuli in each o he expe i-
men s. Mo eo e , mos o hese s udies ocused on nouns,
ypically disyllabic, wi hou cli ics o in lec ions—a pa ial
se o s imuli which do no ep esen he a ie y o wo ds in
he Heb ew language.
In he same ein, ex ensi e wo k has ocused on how
he mo phological s uc u e o Heb ew wo ds a ec s isual
wo d ecogni ion (e.g., Deu sch e al., 1998; Feldman e al.,
1995; F os e al., 1997, 2000). O e all, hese s udies sug-
ges ed ha he oo consonan s a e he co e a ge o wo d
ecogni ion, and ha lexical o ganiza ion in Heb ew ollows
mo phological p inciples. Fo example, F os and colleagues
(2005) showed ha in con as o English, F ench, o Span-
ish, ull o hog aphic o e lap be ween p imes and a ge s
in Heb ew esul s in e y weak masked o hog aphic p im-
ing, in e p e ed as e idence ha he lexical a chi ec u e o
Heb ew p obably does no align, s o e, o connec wo ds
by i ue o hei ull sequence o le e s. Indeed, consid-
e ing he o e all body o esea ch using masked p iming
in Semi ic languages, eliable acili a ion is consis en ly
ob ained whene e p imes consis o he oo le e s, i e-
spec i e o wha he o he le e s a e (e.g., F os e al., 1997,
2000; Velan e al., 2005; Pe ea e al., 2010, bu see Pe ea
e al., 2014, o signi ican o m p iming e ec s in A abic).
Ano he impo an inding conce ns le e posi ion lex-
ibili y. In Indo-Eu opean languages, dis up ing he o de
o he le e s wi hin a wo d has li le impac on eade s’
abili y o ecognize and ead i co ec ly (e.g., Duñabei ia
e al., 2007; Pe ea & Ca ei as, 2006a, 2006b, 2008; Pe ea
& Lupke , 2003, 2004; Schoonbae & G ainge , 2004).
In con as , Heb ew eade s e eal ex eme le e posi ion
igidi y, and eading is signi ican ly impai ed when wo ds
in ol e ansposed le e s (Velan & F os , 2007, 2009, 2011;
and see F iedmann & G ion, 2001, 2005, o le e posi ion
dyslexia o Heb ew eade s). This c oss-linguis ic di e ence
in le e posi ion lexibili y again e lec s he mo phologi-
cal s uc u e o Heb ew: since many Heb ew oo s sha e a
subse o le e s bu di e in hei o de (e.g., Z.M.R., “ o

8766 Beha io Resea ch Me hods (2024) 56:8761–8783
sing”; R.M.Z., “ o hin ”; Z.R.M., “ o low”), le e posi-
ion in Heb ew mus be igid a he han lexible in o de
o access he co ec oo (see Le ne e al., 2014, o com-
pu a ional e idence, and he ecen PONG model, Snell, in
p ess). Howe e , o complica e hings e e u he , no all
Heb ew wo ds ha e a Semi ic s uc u e. Many wo ds om
a ious o igins (e.g., G eek, Pe sian, English) ha e pe me-
a ed Heb ew h oughou his o y, and a e no oo -de i ed
(and see a simila s a e o a ai s o Mal ese, e.g., Gea y &
Ussishkin, 2018). Indeed, hese wo ds ha e been shown o
be p ocessed di e en ly om Semi ic Heb ew wo ds (Bi an
e al., 2020; Haddad e al., 2018; Velan & F os , 2011; Velan
e al., 2013). Howe e , he dis ibu ional p ope ies o such
non-Semi ic Heb ew wo ds a e no ye known, and will be
examined in his wo k.
The cu en s udy: The Heb ew Lexicon P ojec
Since he Heb ew w i ing sys em ep esen s a s a k con as
o o he alphabe ic o hog aphies, and because wo d ecog-
ni ion expe imen s in Heb ew ha e e ealed di e gen ind-
ings wi h impo an c oss-linguis ic implica ions, con as ing
Heb ew eading pe o mance wi h o he languages p omises
o p o ide impo an insigh s o eading esea ch. Hence, a
da abase o beha io al da a on Heb ew wo ds has a - each-
ing implica ions. He e we p esen an open da a sou ce on
Heb ew isual wo d ecogni ion, he Heb ew Lexicon P ojec
(HeLP), which is he i s o examine p in ed wo d ecog-
ni ion in a Semi ic language on a la ge scale. The p ojec
epo s da a om wo asks: LD and naming. I assembles LD
esponses o 10,000 wo ds and nonwo ds, wi h 5000 o he
wo ds also ha ing addi ional associa ed naming da a. Impo -
an ly, HeLP employs an ecologically alid se o Heb ew
wo ds sampled om a na u al Heb ew p in ed co pus (see
Me hods), so ha all ypes o wo ds a e included, includ-
ing wo ds wi h cli ics, p e ixes, and su ixes, Semi ic and
non-Semi ic, in lec ed and de i ed—a a ie y ha e lec s he
wo ds Heb ew eade s encoun e in hei daily li es.
In line wi h p e ious mega-s udies and open-science
s udies mo e b oadly, HeLP is mean o enable esea ch-
e s o ackle a la ge numbe o ques ions ega ding wo d
ecogni ion in Heb ew and i s simila i ies and di e ences
o o he languages. The goal o his i s pape , o cou se,
is no o co e all such po en ial explo a ions. Ra he , in he
cu en pape , we demons a e he u ili y o he HeLP da a
by add essing a se ies o ounda ional ques ions ela ed, on
he one hand, o he s uc u e o he Heb ew w i ing sys em,
and on he o he o he p edic o s o wo d ecogni ion in
ha language. In pa icula , as de ailed below, ou analyses
examine he p e alence and impac o phonological unce -
ain y and homog aphy in LD and naming asks, he dis-
ibu ions o wo d leng hs and neighbo hood densi ies and
hei impac in a oo -based o hog aphy, and he dis ibu ion
and beha io al consequences o Semi ic and non-Semi ic
s uc u e, as well as he numbe o cli ics. Toge he , hese
analyses esul in mapping, on a la ge scale, he beha io al
impac o gene al p edic o s hough o e lec wo d ecog-
ni ion ac oss languages (lexicali y, equency, wo d leng h,
o hog aphic neighbo hood densi y), bu also p edic o s
ha a e ele an speci ically o eading in Heb ew (Semi ic
s uc u e, p esence o cli ics, and phonological ambigui y).
The s uc u e o he pape is as ollows: Fi s , we p esen
esul s con i ming he eliabili y o he HeLP da a, using
spli -hal es ima es (a bo h he i em and pa icipan le el),
mean o ensu e ha he collec ed da a a e o su icien qual-
i y o subsequen analyses. We hen in oduce desc ip i e
s a is ics o he da a, and p esen he basic e ec s e ealed
in he LD and he naming asks pe aining o bo h gene al
and Heb ew-speci ic p edic o s, as e iewed abo e. Finally,
he implica ions o he esul s a e discussed.
Me hods
Lexical decision ask
S imuli
The p ojec assembled LD da a o 10,000 Heb ew wo ds and
10,000 nonwo ds. Wo ds we e sampled om he Heb ew po -
ion o he Subs2 ec co pus ( an Pa idon & Thompson, 2021)
which has 170 million okens. Fi s , a lis o he 50,000 mos
equen Heb ew wo ds was ex ac ed om he co pus, and all
p ope names and misspelled wo ds we e manually emo ed
om ha lis . We hen andomly sampled 2500 wo ds om
he 5000 mos equen wo ds in he lis , and 7500 wo ds om
he emainde o he equency ange o he il e ed lis , which
oge he made he 10,000 a ge s o he LD ask.
Nonwo ds we e gene a ed by shu ling le e s o all wo ds
om he 50,000-wo d il e ed lis . Wo ds wi h ou o ewe
le e s had all hei le e s eshu led. Wo ds wi h i e o
mo e le e s had hei beginning, middle, o end shu led
andomly ( i s , middle, o las le e s o he wo d). The num-
be o le e s shu led anged om 3 o n − 1. Following his
p ocedu e, 10,000 nonwo ds we e selec ed andomly and
hen inspec ed manually o ensu e hey uly had no meaning
in Heb ew. The shu ling o di e en numbe s o le e s a
di e en loca ions wi hin wo ds a oided manual decisions
ha may in oduce bias, and was mean o p o ide a a ie y
o i ems which could be mined o s udy he de e minan s
o ease o di icul y in ejec ing le e s ings as po en ial
Heb ew wo ds. As such, ou “shu ling” app oach esul ed,
o example, in i ems ha ha e a clea e expec ed p onuncia-
ion (e.g., ביגלרת, טגל), along wi h o he s ha a e mo e
phonologically ambiguous (e.g., רמלש); in nonwo ds wi h
a pseudo-mo phological Semi ic s uc u e (e.g., האחבטה,
8767Beha io Resea ch Me hods (2024) 56:8761–8783
להגהר); and in nonwo ds ha a y in big am equency.
In his i s pape we do no analyze p edic o s o esponses
o nonwo ds, bu he nonwo d da a a e made ully a ailable
wi h he es o he da ase o u u e esea ch.
Twen y sublis s, each comp ising 500 wo ds and 500
nonwo ds (1000 a ge s o e all), we e c ea ed o se e as
s imuli o each expe imen al session in he LD ask. To
ensu e ha each sublis included wo ds om he ull e-
quency ange, he i s wo d was assigned o he i s sublis ,
he second wo d o he second sublis , e c., and he 21s wo d
was assigned again o he i s sublis and so on. Resul s om
he 20 sublis s we e analyzed o examine he gene al e ec s
such as lexicali y (see Supplemen a y Ma e ials S1). Since
Heb ew-speci ic p edic o s equi ed manual coding, only
wo ds om he 10 sublis s ha we e also employed in he
naming ask we e used in analyses conside ing hese p edic-
o s (see de ails on naming ask s imuli, below). Nonwo ds
we e andomly assigned o he sublis s, and Welch - es s
ensu ed ha he wo ds and he nonwo ds in e e y sublis did
no di e signi ican ly in e ms o leng h ( (19,970) = − 0.4,
p = 0.68; see Fig.1A). In con as , wo ds and nonwo ds did
di e in hei mean o hog aphic Le ensh ein dis ance 20
(OLD20), a measu e o o hog aphic neighbo hood densi y
de ined as he mean Le ensh ein dis ance o he 20 closes
o hog aphic neighbo s o an i em (Ya koni e al., 2008),
wi h a mean OLD20 o 1.65 o wo ds and 2.17 o non-
wo ds ( (18,909) = − 74.5, p < 0.001, see Fig.1B).This is
expec ed because wo ds sha e mo e o hog aphic esem-
blance o o he wo ds, whe eas nonwo ds ( esul ing om
shu ling le e s) ha e less esemblance o exis ing wo ds.
Pa icipan s
A o al o 273 pa icipan s comple ed he expe imen al ses-
sions. Da a o nine pa icipan s we e excluded om he
analysis due o echnical di icul ies and poo pe o mance
(mean accu acy lowe han 75%, o mo e han 12% o
esponses unde 300ms). O e all, he esul s o 264 pa ici-
pan s (201 emale) had alid LD da a. Due o manual coding
o s imuli om only a pa ial se o o e lapping lis s ha
we e employed in bo h he LD and naming asks, ou cen-
al models which include Heb ew-speci ic p edic o s we e
conduc ed on da a om 178 pa icipan s. Pa icipan s we e
s uden s a he Heb ew Uni e si y o Je usalem and we e
ec ui ed using he Psychology Depa men ’s pa icipan
ec ui men pla o m. The a e age age o he pa icipan s
was 24.14yea s (SD = 3.53yea s). All pa icipan s decla ed
ha hey had no a en ion o eading disabili ies, ha hei
i s language was Heb ew, and ha hey had no mal o co -
ec ed- o-no mal ision. Pa icipan s ecei ed c edi o pay-
men o hei pa icipa ion a e each expe imen al session.
Each pa icipan could ake pa in as many expe imen al
sessions as hey wished, up o 20 (see below), bu hey we e
equi ed o ake a leas a 15-min b eak be ween sessions,
and could no ake pa in mo e han wo sessions a day.
P ocedu e
In he LD ask, all expe imen al sessions we e pe o med
online om home (using lap op o desk op compu e s only,
i.e., no ia sma phones o able s). The LD ask was buil
using PsychoPy (Pei ce e al., 2019), e sion 2021.2.3, and
hos ed on Pa lo ia. A e signing a consen o m and con-
i ming eligibili y c i e ia, pa icipan s we e ins uc ed ha
in each ial hey would be p esen ed wi h a le e s ing on
he compu e sc een, o which hey had o espond as ap-
idly and as accu a ely as possible as o whe he i o med an
exis ing Heb ew wo d (p essing he “L” key) o no (p ess-
ing he “S” key). The session began wi h 10 p ac ice i-
als consis ing o i e wo ds and i e nonwo ds, ollowed
by he expe imen al s imuli. Each s imulus emained on he
sc een un il a esponse was eco ded, wi h a imeou o wo
Fig. 1 P ope ies o wo ds and nonwo ds in he LD ask. A Dis ibu ion o leng h (in le e s). B Dis ibu ion o OLD20
8768 Beha io Resea ch Me hods (2024) 56:8761–8783
seconds, and a blank sc een o 500ms be ween s imuli.
Ta ge s we e p esen ed in he middle o he sc een, in whi e
on on a da k g ay backg ound, wi h hei size se o 10% o
a pa icipan ’s sc een ( he absolu e dimensions a ied gi en
he online na u e o he ask). The e we e h ee b eaks du -
ing he session, a e 250, 500, and 750 ials. Each expe i-
men al session ypically las ed be ween 20 and 30min.
Naming ask
S imuli
F om he 20 sublis s (i.e., 10,000 wo ds) ha we e used in
he LD ask, we sampled 10 sublis s, wi h a o al o 5000
wo ds o se e as s imuli in he naming ask, main aining
he same equency dis ibu ion as in he ull se o 10,000
wo ds. The 5000 wo ds we e edi ided in o six sublis s
o he naming expe imen , ou o hem wi h 800 wo ds
and wo wi h 900 wo ds. Each o he six naming sublis s
included wo ds om he ull equency ange as desc ibed
abo e (25% om he 5000 mos equen wo ds, and 75%
om he en i e ange o equencies ollowing he 5000 mos
equen wo ds).
Pa icipan s
A o al o addi ional 151 pa icipan s (101 emale) com-
ple ed he naming expe imen al sessions, using he same
ec ui men p ocedu e. The mean age o pa icipan s was
24.6yea s (SD = 4.4yea s). As in he LD ask, pa icipan s
could ake pa in as many expe imen al sessions as hey
wished (up o six, he numbe o sublis s).
P ocedu e
In con as o he web-based LD ask, all expe imen al ses-
sions in he naming ask we e pe o med in he labo a o y.
The naming ask was buil using he Neu oBeha io al Sys-
ems so wa e, P esen a ion, e sion 23.010.27.21. Pa ici-
pan s sa in on o a compu e sc een in a quie expe i-
men oom, wea ing a headse . They we e old ha Heb ew
wo ds would appea on he sc een, one a a ime, and hey
should ead e e y wo d aloud as as and as accu a ely as
hey could. Be o e each wo d, a ixa ion c oss appea ed in
he middle o he sc een o 800ms. The wo d disappea ed
om he sc een once a pa icipan ini ia ed a oice key wi h
a spoken esponse o a e a imeou o 1.5s. The expe i-
men s a ed wi h 10 p ac ice ials, o ensu e ha he oice
key ope a ed co ec ly (expe imen e s adjus ed he h esh-
old i needed). The e we e wo b eaks du ing he expe i-
men , a e 250 and 500 wo ds in he 800-wo d sessions,
and a e 300 and 600 wo ds in he 900-wo d sessions.
Wo ds we e p esen ed in he middle o he sc een, in 55-p .
whi e on on a da k g ay backg ound, aking abou 10% o
he e ical dimension o he sc een. RTs we e measu ed
om he appea ance o a wo d on he sc een o he ac i a-
ion o he oice key. Simila o he LD ask, pa icipan s
who wished o pa icipa e in mul iple sessions had o ake
a b eak o a leas 15min be ween sessions, and could no
pa icipa e in mo e han wo sessions in one day. Each
expe imen al session ypically las ed a ound 30min.
Responses in he naming ask we e eco ded and we e
la e coded by a eam o i e ained esea ch assis an s.
Each esponse was coded bo h o i s accu acy (co ec /
inco ec ; and in a e cases, “unclea ”—see below) and o
he alidi y o he RT da a (i.e., whe he he i s eco ded
audi o y signal, which igge ed he oice key, should
be used in RT analysis). A esponse could be coded as
co ec /inco ec while s ill no ha ing alid associa ed
RT: coding as “in alid RT” was au oma ically assumed
in cases whe e he esponses we e as e han 200ms, as
well as in addi ional cases whe e he code no iced ano he
(in alid) esponse ha igge ed he oice key (e.g., in
cases whe e he e was an ini ial sound such as a cough
o mu mu be o e he pa icipan ead he a ge ). In such
cases, we used he co ec ness da a bu no he RT da a in
he analyses below. A o al o 94.5% o esponses we e
coded as ha ing a alid RT associa ed wi h hem, whe eas
99.4% o esponses had a co ec /inco ec coding (i.e.,
only 0.6% o esponses had “unclea ” as he coding o
co ec ness; in analyses below, we ea hese as NA in
all models).
To ensu e he eliabili y o he coding o naming
esponses, we u he andomly sampled six sessions om
six pa icipan s, wi h 5000 naming ials in o al ( ou
pa icipan s had 800 ials; wo pa icipan s had 900 i-
als). The e-coding o hese sessions was done by ano he
(blinded) esea ch assis an , who used he same coding
scheme as he o iginal code s. We ound an in e - a e eli-
abili y es ima e o Cohen’s
𝜅=0.62
, a alue ep esen ing
“subs an ial” ag eemen be ween a e s (Landis & Koch,
1977), wi h es ima es in he da a om he six andomly
sampled pa icipan s sepa a ely anging om
𝜅=0.52
o
𝜅=0.85
. These es ima es sugges ha he eliabili y o he
naming esponse coding was subs an ial o e all and a a
minimum mode a e in indi idual pa icipan s.
P edic o s o  isual wo d ecogni ion pe o mance:
Wo d‑le el a iables
Figu e2 summa izes he di e en psycholinguis ic p edic-
o s made a ailable o he di e en se s o s imuli wi h he
cu en elease o he HeLP da a. In wha ollows, we p o-
ide mo e in o ma ion abou hese p edic o s.
8769Beha io Resea ch Me hods (2024) 56:8761–8783
Gene al isual wo d ecogni ion p edic o s
The models below use as p edic o s he ollowing gene al
a iables, which a e known o impac wo d ecogni ion
ac oss many languages: (a) lexicali y, whe he a a ge is
a wo d o a nonwo d (in he LD ask); (b) wo d equency
(s ing equency o he su ace o m, log- ans o med),
based on he Subs2 ec co pus ( an Pa idon & Thompson,
2021); (c) wo d leng h (in numbe o le e s); and (d)
OLD20, compu ed using he il e ed lis o 50,000 wo ds
om he Subs2 ec co pus, examining o each i em he
numbe o subs i u ion, inse ion, o dele ion ope a ions
equi ed o u n ha i em in o any o he o he 50,000 wo ds
in he lis . Then, OLD20 was de ined as he mean numbe o
al e a ions in he 20 wo ds ha equi ed he minimal numbe
o al e a ions (Ya koni e al., 2008).
Heb ew‑speci ic p edic o s
In addi ion o hese gene al p edic o s, ou models conside
p edic o s po en ially ele an speci ically o wo d ecogni-
ion in Heb ew (and o he Semi ic languages), gi en p e-
ious esea ch and conside ing he p ope ies o he w i -
ing sys em. As de ailed below, ob aining hese measu es
equi ed excessi e manual coding o esponses. Hence, we
ocused on he 5000 wo ds ha we e used as s imuli in bo h
he naming and he LD asks ( a he han on he ull se o
10,000 wo ds in he LD ask).
Semi ic s uc u e. As e iewed in he In oduc ion, an
impo an ea u e o Heb ew is ha while many wo ds ha e
a Semi ic s uc u e and comp ise oo and wo d pa e n
combina ions, he e a e also many wo ds wi hou Semi ic
s uc u e (see Velan & F os , 2011), which ha e been assimi-
la ed in o Heb ew h oughou his o y. The Semi ic agging
o wo ds was manually pe o med such ha each wo d was
agged in o one o ou ca ego ies: “clea ly Semi ic”, “clea ly
non-Semi ic”, “unde e mined”, o “o he ”. Wo ds we e classi-
ied as “clea ly Semi ic” i hey had an unequi ocal oo which
was p oduc i e and used in di e en phonological pa e ns,
c ea ing a a ie y o wo ds wi h dis inc meanings. Fo exam-
ple, he wo d "כתבתי" (KTBTI, meaning I w o e) is con-
s uc ed om he oo le e s “כ,ת,ב” (K.T.B.), and appea s in
many Heb ew pa e ns o c ea e dis inc wo ds (e.g., "מכתב",
MKTB, /mix a /—a le e ; "מכתיב", MKTIB, /max i /—
dic a es; "כתב", KTB, /ka a /—he w o e; see, e.g., F os
e al., 1997). Wo ds we e agged as clea ly non-Semi ic i hey
could no be decomposed in o a p oduc i e oo and a wo d
pa e n (e.g., he wo d "לימון", LIMWN, /limon/, meaning a
lemon, which was assimila ed in o Heb ew and is no de i ed
om a p oduc i e oo o pa e n). The “unde e mined” ca -
ego y was used o wo ds ha could be ead o analyzed in
mo e han one way (e.g., "אטום", ATWM, can be ead as /
a om/, meaning an a om, a non-Semi ic wo d, o as /a um/
meaning “sealed”, de i ed om he oo A.T.M.), o wo ds
ha could no be unequi ocally classi ied as ha ing a Semi ic
s uc u e. In he “o he ” ca ego y he e we e p eposi ions,
ad e bs, and p onouns, which do no ollow ei he a Semi ic
o non-Semi ic s uc u e. Ou agging e ealed ha 75% o
he 5000 wo ds we e clea ly Semi ic, 18.9% we e de ined
as clea ly non-Semi ic, and 6.1% we e unde e mined/o he
(Fig.3A). Only wo ds wi h a clea Semi ic o non-Semi ic
s uc u e we e included in he models below.
Fig. 2 In o ma ion abou included HeLP s imuli and hei a ailable lexical p ope ies
8776 Beha io Resea ch Me hods (2024) 56:8761–8783
Pa icipan s again showed he expec ed equency e ec
(Z = 15.67, p < 0.001). Howe e , in con as o he LD
esul s, hey made ewe naming e o s in sho e wo ds
(Z = − 2.75, p = 0.006; al hough, when emo ing OLD20
om he model, he leng h e ec lipped, see Supplemen a y
Ma e ials S2). The e was a signi ican OLD20 e ec , indi-
ca ing ha pa icipan s we e mo e accu a e when a wo d had
ewe o hog aphic neighbo s (Z = 7.26, p < 0.001). In e ms
o in e ac ions (Fig.9), he e was a signi ican in e ac ion
be ween wo d leng h and OLD20 (Z = − 6.32, p < 0.001):
pa icipan s made mo e e o s naming sho wo ds wi h
many o hog aphic neighbo s, and made mo e e o s nam-
ing longe wo ds wi h ew o hog aphic neighbo s. The e
was also a signi ican in e ac ion be ween log equency
and wo d leng h (Z = − 3.02, p = 0.003), wi h leng h e ec s
e ealed only o wo ds in he mid- equency ange and
abo e (bu no in low- equency wo ds).
Fig. 7 Visual depic ion o signi ican In e ac ions in he accu acy
model, LD da a. A In e ac ion be ween wo d leng h and wo d log
equency. B In e ac ion be ween OLD20 and wo d log equency. C
In e ac ion be ween OLD20 and wo d leng h
Fig. 8 Visual depic ion o e ec s o in e es in he RT model, naming
da a. A In e ac ion be ween OLD20 and wo d leng h ( he only sig-
ni ican in e ac ion in his model). B Es ima ed mean log- ans o med
RT o wo ds as a unc ion o he numbe o cli ics

8777Beha io Resea ch Me hods (2024) 56:8761–8783
As o Heb ew-speci ic p edic o s, he e was no di e ence
in naming accu acy be ween Semi ic and non-Semi ic wo ds
(Z = 0.63, p = 0.53). Pa icipan s made mo e e o s eading
wo ds wi h one (Z = − 4.75, p < 0.001), wo (Z = − 2.87,
p = 0.004), and h ee cli ics (Z = − 3.29, p < 0.001) han
wo ds wi hou cli ics. Also, as expec ed, pa icipan s made
mo e e o s eading wo ds wi h highe p onuncia ion
en opy (Z = − 3.71, p < 0.001).
Discussion
How he basic cha ac e is ics o w i ing sys ems impac
isual wo d ecogni ion beha io has been he ocus o
ex ensi e esea ch (see, e.g., F os , 2012, o e iew and
discussion). In his ein, s udies conduc ed in di e en
languages, pa icula ly hose including la ge-scale LPs,
ha e p o ided impo an insigh s ega ding he di e en
compu a ions eade s employ du ing eading in hei w i -
ing sys em, highligh ing high-o de p inciples o wo d ec-
ogni ion. F om his pe spec i e, e idence om Heb ew has
con inuously shaped heo ies and models o eading, gi en
he unique cha ac e is ics o i s w i ing sys em. Howe e ,
wo d ecogni ion s udies in Heb ew o da e ha e employed
small, a ge ed expe imen s, co e ing only a limi ed pa o
he language’s lexicon. To add ess his gap, we p esen he e
o he i s ime a la ge-scale da ase o eading beha io in
a Semi ic language, Heb ew, comp ising LD esponses o
10,000 wo ds and nonwo ds, and naming esponses o 5000
wo ds. In his i s pape , we hen u ilize he da a om he
Heb ew LP (HeLP) o examine he con ibu ion o gene al
p edic o s (lexicali y, equency, leng h, and o hog aphic
neighbo hood), and Heb ew-speci ic p edic o s (Semi ic
s uc u e, cli ic le e s, and ex en o phonological ambigu-
i y), o isual wo d ecogni ion pe o mance. As we discuss
below in de ail, ou indings o e impo an insigh s ega d-
ing he compu a ions in ol ed in he p ocessing o p in ed
wo ds in a w i ing sys em such as Heb ew, sugges ing a se
o uni e sal compu a ions in ol ed in p in p ocessing.
Gene al p edic o s o wo d ecogni ion
Unsu p isingly, he benchma k e ec s o equency and
lexicali y eme ged in Heb ew as in any LP, con i ming ha
hese p inciples o lexical sea ch a e simila ac oss w i ing
sys ems. O heo e ical in e es , he e o e, a e indings in
which Heb ew seems o di e ge om he well- esea ched
Eu opean languages.
A i s inding o in e es is he e ec (o lack he eo ) o
wo d leng h, mainly in LD ( he pa allel e ec s in naming
a e somewha weake ). When in e p e ing he HeLP indings
o leng h, an impo an ac o o conside is he collinea -
i y be ween wo d leng h and OLD20 ( = 0.79). Tha sho e
wo ds ha e mo e o hog aphic neighbo s is ypical o many
w i ing sys ems, as e ealed in o he LPs (e.g., a co ela ion
o 0.77 in F ench, Fe and e al., 2010; a co ela ion o 0.79
in Eu opean Po uguese, Soa es e al., 2019). Bu no e ha in
he as majo i y o LPs, leng h e ec s we e signi ican e en
when his high co ela ion was pa ialed ou in he analyses.
Fo example, in he Malay LP, he se e al leng h measu es
ha we e used showed co ela ions anging om 0.47 o 0.77
wi h he OLD20 sco es, and ye , independen ly, wo d leng h
was he s onges p edic o o pe o mance in bo h LD and
naming asks (Yap e al., 2010). In ligh o hose p e ious
esul s, he consis en lack o wo d leng h e ec o wo ds
in Heb ew s ands ou , and models ha also included OLD20
o en e ealed a e e sed leng h e ec , whe e longe wo ds
incu ed as e ecogni ion ime and g ea e accu acy. Com-
pa ed wi h o he exis ing LPs, hese esul s seem o align
Fig. 9 Visual depic ion o signi ican In e ac ions in he accu acy
model, naming da a. A In e ac ion be ween wo d leng h and wo d log
equency. B In e ac ion be ween OLD20 and wo d leng h
8778 Beha io Resea ch Me hods (2024) 56:8761–8783
wi h indings in he Pe sian LP (Nema i e al., 2022). A pos-
sible common ac o o Heb ew and Pe sian is ha in bo h
languages, mos owels a e omi ed om o hog aphic ep-
esen a ion o wo ds. Since longe wo ds on a e age include
mo e owel le e s, hey incu less phonological ambigui y,
poin ing o a lexical candida e mo e apidly and esul ing in
as e ecogni ion.4 Mo e b oadly, ou indings esona e wi h
p e ious claims ha wo d leng h e ec s a e s onge in mo e
anspa en w i ing sys ems (e.g., Cue os & Sua ez-Coalla,
2009; Ellis & Hoope , 2001; and see Weiss e al., 2015, o
ela ed e idence om poin ed s. un-poin ed Heb ew). Ou
esul s go a s ep u he o sugges ha in Heb ew, wo d
leng h e ec s a e some imes e e sed, a guably due o lowe
le els o phonological ambigui y in longe wo ds.
A second poin o in e es is he di e en impac o o ho-
g aphic neighbo s in he LD e sus he naming ask. While
he ypical acili a o y e ec o o hog aphic neighbo hood
was obse ed in LD, ou indings show ei he ha o ho-
g aphic neighbo hood was no p edic i e o naming RT and
accu acy, o ha ha ing many o hog aphic neighbo s o a
wo d esul ed in an inhibi o y e ec . This con as s wi h s ud-
ies showing ha he p esence o many o hog aphic neigh-
bo s has a acili a o y e ec on naming English wo ds (e.g.,
Ya koni e al., 2008). Impo an ly, he e a e documen ed
c oss-linguis ic di e ences in o hog aphic neighbo hood
e ec s in naming. Fo example, simila o ou p esen inding
in Heb ew, Chang e al. (2016) epo ed an inhibi o y e ec
o neighbo hood size in Chinese, and demons a ed h ough
compu a ional modeling ha he di ision o labo be ween
phonological and seman ic pa hways in deep o hog aphies
is he key o accoun ing o he inhibi o y e ec o neighbo -
hood size (and see Pee eman & Con en , 1997, o di e ences
be ween English and F ench). The di e ence be ween LD and
naming, hen, e lec s subs an ial di e ences in compu a ions
in a deep o hog aphy like Heb ew. While he p esence o
many o hog aphic neighbo s could con ibu e o a as deci-
sion o whe he a le e s ing is a Heb ew wo d o no , naming
equi es one o iden i y, selec , and p onounce speci ic lexical
candida es ha o en di e in owel con igu a ions. Hence,
o Heb ew, he p esence o many compe i o s seems o slow
esponse imes, a he han accele a e hem.
Heb ew‑speci ic e ec s
Gi en he unique p ope ies o Heb ew, unde s anding he
ole o he Heb ew-speci ic p edic o s p o ides impo an
insigh s wi h ega d o isual wo d ecogni ion in ha lan-
guage. Hence, in his sec ion we e iew indings pe ain-
ing o he h ee Heb ew-speci ic p edic o s es ed: Semi ic
s uc u e, p esence o cli ics, and phonological ambigui y.
Wi h ega d o Semi ic s uc u e, e en hough he Semi ic
agging we employed was conse a i e (de ining a wo d as
“clea ly Semi ic” only i i s oo le e s we e clea ly p oduc-
i e), 75% o he 5000 wo ds we e agged as such. Mos o
ou s a is ical models showed ha pa icipan s’ pe o mance
imp o ed when p esen ed wi h wo ds ha ing a Semi ic
s uc u e. The p esen indings o HeLP indica e hen ha
Heb ew eade s become a uned o he s a is ical p ope ies
o Heb ew wi h i s non-conca ena ed mo phology, and a e
mo e e icien in p ocessing Heb ew wo ds when hey con-
o m o he highly p e alen Semi ic o m o in e wined
oo and wo d pa e n mo phemes. We assume ha h ough
s a is ical lea ning, eade s become inc easingly e icien in
de ec ing he oo le e s wi hin p in ed Semi ic wo ds, ena-
bling he as decomposi ion o p in ed wo ds in o hei con-
s i uen mo phemes (see, e.g., Feldman e al., 1995; Velan
e al., 2013). This p o ides eade s wi h he missing owel
in o ma ion, and leads o as lexical access when wo ds a e
o ganized by mo phological a he han simple o hog aphic
p inciples (see F os e al., 2005, o discussion).
Rega ding he e ec s o cli ics, ou indings indica e ha
he numbe o cli ic le e s impac ed pe o mance, wi h
esponses o wo ds wi h cli ics gene ally being less accu a e
and incu ing slowe RTs. This inding sugges s ha when
wo ds appea in isola ion wi hou disambigua ing con ex ,
cli ic le e s add u he complexi y o he p ocess o decom-
posing he p in ed wo d in o i s mo phemic cons i uen s. An
in e es ing de ia ion om his pa e n, howe e , was ound
in he naming ask: pa icipan s ead aloud wo ds wi h cli -
ics as e (al hough s ill wi h mo e e o s). Gi en he high
p e alence o cli ics and he ela i e sys ema ici y o hei
p onuncia ion (e.g., “and” in mos cases is p onounced as
e , “ he” as ha , e c.), and in ligh o he ime cons ain s
in he naming ask, we canno ule ou he possibili y ha
pa icipan s ini ia ed he p onuncia ion o he ini ial cli ics
be o e ully ecognizing he wo d (and he highe e o a e
o wo ds wi h cli ics indeed suppo s his explana ion). As
naming la encies e lec he ime cou se o he ini ial u e -
ance, ou o e all speeded esponses in he naming ask could
e lec pa icipan s’ high con idence in ini ia ing a ocal
esponse o he ini ial cli ic le e .
The hi d impo an cha ac e is ic o Heb ew is i s pho-
nological unde -speci ica ion. While p e ious esea ch
simply coun ed he numbe o emp y owel slo s o assess
phonological unce ain y (e.g., F os , 1995), in he p esen
4 Some eade s may wonde why he co ela ion be ween wo d
leng h and phonological en opy, while nega i e, is qui e small,
= − 0.16 (Table3 abo e). In his con ex , i is impo an o keep in
mind ha ou measu e o phonological en opy e lec s he numbe
o meaning ul p onuncia ions a wo d has, gi en he he e ophonic
aspec o Heb ew. As such, i does no ully cap u e a wo d’s deg ee
o phonological ambigui y (i.e., unce ain y o inconsis ency): I does
no e lec he di icul y in gene a ing unique p onuncia ions gi en he
numbe o missing owels in he o hog aphic sequence. Hence wo
wo ds can ha e a unique lexical p onuncia ion (and hence a simila
sco e in he measu e o phonological en opy), bu one wi h ew miss-
ing owels and one wi h many.
8779Beha io Resea ch Me hods (2024) 56:8761–8783
wo k we cap u ed he le el o phonological unce ain y
when p onouncing a wo d by ma hema ically quan i ying
phonological en opy ac oss esponses in he ull sample o
pa icipan s (see also De Simone e al., 2021). Ou esul s
show ha in all models, highe p onuncia ion en opy hin-
de ed pe o mance, o bo h naming and LD. Whe eas he
impac o phonological en opy in naming is any hing bu
su p ising, he pa allel inding o LD is s iking. The o ho-
g aphical dep h hypo hesis (F os e al., 1987) has a gued
ha eade s o shallow o hog aphies ely s ongly on pho-
nological cues in isual wo d ecogni ion, while eade s o
deep o hog aphies such as English o Heb ew ely mo e on
o hog aphic o seman ic cues. I was assumed ha in deep
o hog aphies, he phonological in o ma ion o a wo d is
media ed by he in e nal lexicon (F os e al., 1987; Ka z &
Feldman, 1981). Ou p esen da a sugges s ha phonological
in o ma ion is compu ed no only when p onouncing a wo d,
bu also when simply iden i ying i in he LD ask. This ind-
ing acco ds wi h he claim ha ea ly and as phonological
compu a ions in isual wo d ecogni ion cha ac e ize he
eading p ocess in any o hog aphy whe he shallow o deep
( o discussion see F os , 1998; Ras le & B ysbae , 2006).
Indeed, Rueckl e al. (2015) ha e shown ha eade s o di -
e en languages wi h di e en o hog aphic dep hs demon-
s a e simila neu onal p ocessing o p in ed wo ds, includ-
ing in b ain a eas associa ed wi h phonological p ocessing.
Ou esul s suppo his line o esea ch, showing ea ly
phonological p ocessing in Heb ew, which is conside ed a
highly deep o hog aphy. Tha said, we should cau ion ha
ou measu e o phonological en opy also po en ially e lec s
unce ain y in he mapping be ween p in and meaning; his
is because, in Heb ew, mul iple p onuncia ions o he same
wo d o m also o en ha e mul iple dis inc meanings, and
he e o e mo e homog aphic wo d o ms also ypically ca y
mo e ambigui y in he o hog aphic–seman ic mapping.
Fu u e wo k should ca e ully disen angle he e ec s o di -
e en ypes o ambigui y by p o iding and alida ing wo d-
le el measu es o p in –speech and p in –meaning egula i-
ies in Heb ew. We expec he HeLP da a o be c ucial in he
alida ion o such measu es ( o pa allel wo k in English
using he ELP da a, see, e.g., Chee e al., 2020; Ma elli &
Amen a, 2018; Siegelman e al., 2020, 2022).
Wha s ands ou ? Wha isuni e sal?
As ou lined in ou In oduc ion, ou heo e ical app oach o
LPs is ha hey go a beyond he desc ip i e s a is ics o
ye ano he w i ing sys em. Di e gen indings in LPs poin
o highe -o de compu a ional p inciples ha explica e he
di e ence in esul s in one language ela i e o ano he . He e
we a gue ha o accoun o he ange o indings e ealed
in HeLP, isual wo d ecogni ion should be conside ed as a
p ocess o unce ain y educ ion wi h espec o he iden i y
o lexical candida es and hei phonological s uc u e, as
ep esen ed by he p in ed o ms. This uni e sal p inciple
accoun s o c oss-linguis ic di e ences by weigh ing he
se o cons ain s ha d i e unce ain y educ ion in a gi en
w i ing sys em.
The main p oblem in eading Heb ew is in con e ging
on an unequi ocal lexical and phonological solu ion o a
ange o pa sing and decoding al e na i es. While eading
in con ex signi ican ly educes unce ain y, leading in mos
cases o a single solu ion, wo ds in isola ion incu signi ican
unce ain y. Unce ain y in Semi ic languages conce ns com-
pe ing pa sing possibili ies (e.g., whe he he ini ial le e is
a cli ic le e , a wo d pa e n le e , o a oo le e , which is
c i ical o iden i ying he co ec lemma) and also compe -
ing phonological ep esen a ions. This pe spec i e accoun s,
o example, o he as e esponses o longe wo ds, since
o en hey incu lowe en opy han sho e wo ds. I o e s
a possible explana ion o he as e esponses o wo ds wi h
Semi ic s uc u e, since hese wo ds ypically con ain se e al
cues o co ec mo phological decomposi ion (and see Ba -
On e al., 2017, 2019, 2021, o discussions o unce ain y
educ ion in Heb ew).
Conside ing isual wo d ecogni ion as a p ocess o
unce ain y educ ion also ou lines he ange o dimensions
o conside when compa ing pe o mance ac oss languages,
and gene a es p edic ions ega ding c oss-linguis ic di e -
ences in eading. I shi s he scope o analysis om uni-
dimensional ac o s such as p in –speech anspa ency, o
mo phological complexi y, o ega d pe o mance in isual
wo d ecogni ion in e ms o cons ain sa is ac ion, whe e
mul idimensional cons ain s in e ac o de e mine he ou -
come o p ocessing.
Fu u e di ec ions
Ou p esen indings o e compelling e idence o how
he unique mo phological and phonological p ope ies o
Heb ew play an impo an ole du ing isual wo d ecogni-
ion. Howe e , ou p esen analyses a e bu a i s s ep which
in ol es coa se-g ained quan i ica ion o wo ds’ p ope ies.
Following in he s eps o p e ious LPs, he HeLP p ojec
adhe es o he p inciples o open science, making all da a
a ailable o seconda y analyses, which we hope will acili-
a e u u e in es iga ion in o he exac p edic o s o isual
wo d ecogni ion in Heb ew.
As men ioned b ie ly abo e, one impo an a enue o
u u e esea ch is he de elopmen o mo e sub le and p e-
cise quan i ica ion o o hog aphic–phonological egula i ies
in Heb ew. In he cu en wo k, we only used a measu e o
p onuncia ion en opy, which was calcula ed gi en he ac ual
p onuncia ions ha pa icipan s u e ed in he naming ask.
Al hough his measu e has ecological alidi y as i includes
all p onuncia ions ha we e exp essed in p ac ice by ou
8780 Beha io Resea ch Me hods (2024) 56:8761–8783
sample o pa icipan s, i is limi ed in wo impo an ways.
Fi s , he e a e o he phonological exp essions o he wo ds
we employed ha we e no included in he measu e’s calcu-
la ion. Fo example, he p in ed wo d "כמורה" (KMWRH)
has i e di e en phonological o ms ha bea meaning in
Heb ew (/kmu a/, /kemo e/, /kemo a/, /kamo e/, /kamo a/),
bu only h ee o hem we e p oduced by ou pa icipan s.
Pe haps mo e impo an ly, conside ing only ac ual p onun-
cia ions does no cap u e he ull ex en o which di e en
g aphemes p edic phonemes in he language. We lea e i o
u u e wo k o de elop p ecise co pus-based me ics o he
links be ween o hog aphy and phonology in Heb ew. This
wo k will mos likely in ol e adap ing measu es de eloped
in English and o he Eu opean languages (e.g., Chee e al.,
2020; Siegelman e al., 2020) o cap u e he unique p ope -
ies o Heb ew (e.g., he ac ha in Heb ew, in con as o
English, subs an ial i egula i ies also exis in he mapping
o consonan le e s in o phonemes, a he han mos ly owel
le e s).
No e ha p oducing p ecise measu es o o ho-
g aphic–phonological en opy is bu one s ep in assessing
unce ain y in an Abjad w i ing sys em such as Heb ew.
O he a enues would include he impac o phonologi-
cal Le ensh ein dis ance, as well as measu es o o ho-
g aphic–seman ic egula i ies (e.g., Ma elli & Amen a,
2018; Siegelman e al., 2022). Mul iple o he a enues o
analyses using he HeLP da a can eplica e s udies using
he ELP da a in a w i ing sys em wi h a di e gen s uc-
u e, including hose in psycholinguis ic a ings such as
wo d conc e eness, (B ysbae e al., 2014), age o acqui-
si ion (Kupe man e al., 2012), and body–objec in e ac-
ion (Pexman e al., 2019), o name a ew. Fu u e analyses
should also conside he possible in e ac ions ha may
exis be ween Heb ew-speci ic p edic o s and gene al
psycholinguis ic p ope ies (e.g., he ac ha Semi ic
s uc u e o ex en o phonological ambigui y may impac
p ocessing di e en ly ac oss wo d leng hs). Me ging all
hese dimensions oge he would enable he alignmen o
w i ing sys ems o c oss-linguis ic compa isons, = while
simul aneously conside ing he possible in e ac ions o
hei o hog aphic, phonological, and seman ic p ope ies.
Supplemen a y in o ma ion The online e sion con ains supplemen-
a y ma e ial a ailable a h ps:// doi. o g/ 10. 3758/ s13428- 024- 02502-4.
Acknowledgemen s Wo k in his pape was suppo ed by he ollow-
ing unding sou ces: he Eu opean Resea ch Council (ERC) Ad anced
G an , p ojec 692502-L2STAT, unde he Ho izon 2020 esea ch and
inno a ion p og am (awa ded o RF); he Is ael Science Founda ion
(ISF) G an , p ojec 705/20 (awa ded o RF); he Is ael Science Foun-
da ion (ISF) G an , p ojec 1034/23 (awa ded o NS); and an Az ieli
Ea ly Ca ee Facul y Fellowship (awa ded o NS). We hank Gollan
Anko i, Shaked Sukiennik, Eina A aham, Noam Da ido , and Ellah
Rich e o hei wo k on manually coding he Heb ew-speci ic p edic-
o s and help in da a collec ion.
Funding Open access unding p o ided by Heb ew Uni e si y o Je u-
salem. Wo k in his pape was suppo ed by he ollowing unding
sou ces: he Eu opean Resea ch Council (ERC) Ad anced G an , p o-
jec 692502-L2STAT, unde he Ho izon 2020 esea ch and inno a ion
p og am (awa ded o RF); he Is ael Science Founda ion (ISF) G an ,
p ojec 705/20 (awa ded o RF); he Is ael Science Founda ion (ISF)
G an , p ojec 1034/23 (awa ded o NS); and an Az ieli Ea ly Ca ee
Facul y Fellowship (awa ded o NS).
Da a a ailabili y The ull HeLP da a a e a ailable ia he Open Science
F amewo k (OSF) websi e o seconda y da a analyses, along wi h he
ull lis s o i ems used in he s udies: h ps:// os . io/ nxq8g/.
Code a ailabili y The p ojec ’s OSF eposi o y also includes he code
used o analyses epo ed in his pape .
Decla a ions
E hics app o al This s udy was pe o med in line wi h he p inciples o
he Decla a ion o Helsinki. App o al was g an ed by he Ins i u ional
Re iew Boa d o he Heb ew Uni e si y o Je usalem (App o al No.:
17032022, g an ed Ma ch 17 h, 2022).
Consen o pa icipa e In o med consen was ob ained om all indi-
idual pa icipan s included in he s udy by signing a physical o i ual
consen o m.
Consen o publica ion Only de-iden i ied da a a e a ailable in he
p ojec ’s eposi o y. Pa icipan s ag eed o ha ing hei de-iden i ied
da a sha ed as pa o a publica ion.
Con lic s o in e es The au ho s ha e no ele an inancial o non-
inancial in e es s o disclose.
Open Access This a icle is licensed unde a C ea i e Commons A i-
bu ion 4.0 In e na ional License, which pe mi s use, sha ing, adap a-
ion, dis ibu ion and ep oduc ion in any medium o o ma , as long
as you gi e app op ia e c edi o he o iginal au ho (s) and he sou ce,
p o ide a link o he C ea i e Commons licence, and indica e i changes
we e made. The images o o he hi d pa y ma e ial in his a icle a e
included in he a icle’s C ea i e Commons licence, unless indica ed
o he wise in a c edi line o he ma e ial. I ma e ial is no included in
he a icle’s C ea i e Commons licence and you in ended use is no
pe mi ed by s a u o y egula ion o exceeds he pe mi ed use, you will
need o ob ain pe mission di ec ly om he copy igh holde . To iew a
copy o his licence, isi h p://c ea i ecommons.o g/licenses/by/4.0/.
Re e ences
Aguas i as, J. A., Ca ei as, M., B ysbae , M., Mande a, P., Keulee s,
E., & Duñabei ia, J. A. (2018). SPALEX: A Spanish lexical deci-
sion da abase om a massi e online da a collec ion. F on ie s in
Psychology, 9, 2156.
And ews, S. (1992). F equency and neighbo hood e ec s on lexical
access: Lexical simila i y o o hog aphic edundancy? Jou nal
o Expe imen al Psychology: Lea ning, Memo y, and Cogni ion,
18(2), 234.
Balo a, D. A., Yap, M. J., Hu chison, K. A., Co ese, M. J., Kessle , B.,
Lo is, B., ... & T eiman, R. (2007). The English lexicon p ojec .
Beha io Resea ch Me hods, 39, 445–459.
Ba -On, A., Da ne , E., & B aun-Pe e z, O. (2019). Resol ing hom-
og aphy: The ole o pos -homog aph con ex in eading aloud
8781Beha io Resea ch Me hods (2024) 56:8761–8783
ambiguous sen ences in Heb ew. Applied Psycholinguis ics,
40(6), 1405–1420.
Ba -On, A., O on, T., & Peleg, O. (2021). Seman ic and syn ac ic
cons ain s in esol ing homog aphy: A de elopmen al s udy in
Heb ew. Reading and W i ing, 34, 2103–2126.
Ba -On, A., Da ne , E., & Ra id, D. (2017). Con ex e ec s on he -
e ophonic-homog aphy esolu ion in lea ning o ead Heb ew.
Reading and W i ing, 30, 463–487.
Ben in, S., & F os , R. (1987). P ocessing lexical ambigui y and isual
wo d ecogni ion in a deep o hog aphy. Memo y & Cogni ion,
15(1), 13–23.
Bi an, T., Weiss, Y., Ka zi , T., & T uzman, T. (2020). Mo phological
decomposi ion compensa es o impe ec ions in phonological
decoding. Neu al e idence om ypical and dyslexic eade s o
an opaque o hog aphy. Co ex, 130, 172–191.
B oadben , D. E. (1967). Wo d- equency e ec and esponse bias.
Psychological Re iew, 74(1), 1.
B ysbae , M., Lag ou, E., & S e ens, M. (2017). Visual wo d ecog-
ni ion in a second language: A es o he lexical en enchmen
hypo hesis wi h lexical decision imes.Bilingualism: Language
and Cogni ion,20(3), 530–548.
B ysbae , M., Wa ine , A. B., & Kupe man, V. (2014). Conc e eness
a ings o 40 housand gene ally known English wo d lemmas.
Beha io Resea ch Me hods, 46, 904–911.
Chang, Y. N., Welbou ne, S., & Lee, C. Y. (2016). Explo ing o ho-
g aphic neighbo hood size e ec s in a compu a ional model o
Chinese cha ac e naming. Cogni i e Psychology, 91, 1–23.
Chee, Q. W., Chow, K. J., Yap, M. J., & Goh, W. D. (2020). Consis -
ency no ms o 37,677 English wo ds.Beha io Resea ch Me h-
ods,52(6), 2535–2555.s
Cue os, F., & Suá ez-Coalla, P. (2009). F om g apheme o wo d in
eading acquisi ion in Spanish. Applied Psycholinguis ics, 30(4),
583–601.
De Simone, E., Beye smann, E., Mula i, C., Mi aul , J., &
Schmalz, X. (2021). O de among chaos: C oss-linguis ic di -
e ences and de elopmen al ajec o ies in pseudowo d ead-
ing aloud using p onuncia ion En opy. PLoS ONE, 16(5),
e0251629.
Deu sch, A., F os , R., & Fo s e , K. I. (1998). Ve bs and nouns a e
o ganized and accessed di e en ly in he men al lexicon: E i-
dence om Heb ew. Jou nal o Expe imen al Psychology:
Lea ning, Memo y, and Cogni ion, 24(5), 1238.
Deu sch, A., F os , R., Pelleg, S., Polla sek, A., & Rayne , K. (2003).
Ea ly mo phological e ec s in eading: E idence om pa a o-
eal p e iew bene i in Heb ew. Psychonomic Bulle in & Re iew,
10(2), 415–422.
Deu sch, A., Velan, H., Me zbach, Y., & Michaly, T. (2021). The
dependence o oo ex ac ion in a non-conca ena ed mo phology
on he wo d-speci ic o hog aphic con ex . Jou nal o Memo y
and Language, 116, 104182.
Duñabei ia, J. A., Pe ea, M., & Ca ei as, M. (2007). Do ansposed-
le e simila i y e ec s occu a a mo pheme le el? E idence
o Mo pho-O hog aphic Decomposi ion. Cogni ion, 105(3),
691–703.
Dyma ska, A., Connell, L., & Banks, B. (2023). Weake han you migh
imagine: De e mining imageabili y e ec s on wo d ecogni ion.
Jou nal o Memo y and Language, 129, 104398.
Ellis, N. C., & Hoope , A. M. (2001). Why lea ning o ead is easie
in Welsh han in English: O hog aphic anspa ency e ec s
e inced wi h equency-ma ched es s. Applied Psycholinguis-
ics, 22(4), 571–599.
Feldman, L. B., F os , R., & Pnini, T. (1995). Decomposing wo ds
in o hei cons i uen mo phemes: E idence om English and
Heb ew. Jou nal o Expe imen al Psychology: Lea ning, Mem-
o y, and Cogni ion, 21(4), 947.
Fe and, L., New, B., B ysbae , M., Keulee s, E., Bonin, P., Méo , A.,
Augus ino a, M., & Pallie , C. (2010). The F ench lexicon p o-
jec : Lexical decision da a o 38,840 F ench wo ds and 38,840
pseudo wo ds. Beha io Resea ch Me hods, 42(2), 488–496.
F ed iksen, J. R., & K oll, J. F. (1976). Spelling and sound: App oaches
o he in e nal lexicon. Jou nal o Expe imen al Psychology:
Human Pe cep ion and Pe o mance, 2, 361–379.
F iedmann, N., & G ion, A. (2001). Le e posi ion dyslexia. Cogni i e
Neu opsychology, 18(8), 673–696.
F iedmann, N., & G ion, A. (2005). Le e o m as a cons ain o
e o s in neglec dyslexia and le e posi ion dyslexia. Beha -
iou al Neu ology, 16(2–3), 145–158.
F iedmann, N., & Luko , L. (2008). De elopmen al su ace dyslexias.
Co ex, 44(9), 1146–1160.
F iedmann, N., & Rahamim, E. (2007). De elopmen al le e posi ion
dyslexia. Jou nal o Neu opsychology, 1(2), 201–236.
F os , R. (2006). Becoming li e a e in Heb ew: The g ain size hypo h-
esis and Semi ic o hog aphic sys ems. De elopmen al Science,
9(5), 439.
F os , R. (1995). Phonological compu a ion and missing owels: Map-
ping lexical in ol emen in eading. Jou nal o Expe imen al
Psychology: Lea ning, Memo y, and Cogni ion, 21(2), 398.
F os , R. (1994). P elexical and pos lexical s a egies in eading:
E idence om a deep and a shallow o hog aphy. Jou nal o
Expe imen al Psychology: Lea ning, Memo y, and Cogni ion,
20(1), 116.
F os , R. (1998). Towa d a s ong phonological heo y o isual wo d
ecogni ion: T ue issues and alse ails. Psychological Bulle in,
123(1), 71–99.
F os , R. (2012). Towa ds a uni e sal model o eading. Beha io al and
B ain Sciences, 35(5), 263–279.
F os , R., Deu sch, A., Gilboa, O., Tannenbaum, M., & Ma slen-
Wilson, W. (2000). Mo phological p iming: Dissocia ion o
phonological, seman ic, and mo phological ac o s. Memo y &
Cogni ion, 28(8), 1277–1288.
F os , R., Fo s e , K. I., & Deu sch, A. (1997). Wha can we lea n om
he mo phology o Heb ew? A masked-p iming in es iga ion o
mo phological ep esen a ion. Jou nal o Expe imen al Psychol-
ogy: Lea ning Memo y and Cogni ion, 23(4), 829–856.
F os , R., Kugle , T., Deu sch, A., & Fo s e , K. I. (2005). O hog aphic
S uc u e Ve susMo phological S uc u e: P inciples o Lexical
O ganiza ion in a Gi en Language. Jou nal o Expe imen al Psy-
chology: Lea ning, Memo y, and Cogni ion, 31(6), 1293–1326.
F os , R., Ka z, L., & Ben in, S. (1987). S a egies o Visual Wo d
Recogni ion and O hog aphical Dep h: A Mul ilingual Com-
pa ison. Jou nal o Expe imen al Psychology: Human Pe cep ion
and Pe o mance, 13(1), 104–115.
Fo s e , K. I., & Chambe s, S. M. (1973). Lexical access and naming
ime. Jou nal o Ve bal Lea ning and Ve bal Beha io , 12(6),
627–635.
Gea y, J. A., & Ussishkin, A. (2018). Roo -le e p iming in Mal ese
isual wo d ecogni ion. The Men al Lexicon, 13(1), 1–25.
G ainge , J., Ke in O’ egan, J., Jacobs, A. M., & Segui, J. (1989). On
he ole o compe ing wo d uni s in isual wo d ecogni ion:
The neighbo hood equency e ec . Pe cep ion & Psychophys-
ics, 45(3), 189–195.
Haddad, L., Weiss, Y., Ka zi , T., & Bi an, T. (2018). O hog aphic
anspa ency enhances mo phological segmen a ion in child en
eading Heb ew wo ds. F on ie s in Psychology, 8, 2369.
Hi sho n, E. A., & Ha is, L. N. (2022). Cul u e is no des iny, o
eading: Highligh ing a iable ou es o li e acy wi hin w i ing
sys ems. Annals o he New Yo k Academy o Sciences, 1513(1),
31–47.
Hudson, P. T., & Be gman, M. W. (1985). Lexical knowledge in wo d
ecogni ion: Wo d leng h and wo d equency in naming and

8782 Beha io Resea ch Me hods (2024) 56:8761–8783
lexical decision asks. Jou nal o Memo y and Language, 24(1),
46–58.
Ka z, L., & Feldman, L. B. (1981). Linguis ic coding in wo d ecogni-
ion: Compa isons be ween a deep and a shallow o hog aphy. In
A. Lesgold & C. Pe e i (Eds.), In e ac i e P ocesses in Reading
(pp. 85–99). Law ence E lbaum Associa es.
Ka z, L., & F os , R. (1992). The eading p ocess is di e en o di e -
en o hog aphies: The o hog aphic dep h hypo hesis. Ad ances
in Psychology, 94, 67–84.
Keulee s, E., Diependaele, K., & B ysbae , M. (2010). P ac ice e ec s
in la ge-scale isual wo d ecogni ion s udies: A lexical decision
s udy on 14,000 Du ch mono-and disyllabic wo ds and nonwo ds.
F on ie s in Psychology, 1, 174.
Keulee s, E., Lacey, P., Ras le, K., & B ysbae , M. (2012). The B i ish
Lexicon P ojec : Lexical decision da a o 28,730 monosyllabic
and disyllabic English wo ds. Beha io Resea ch Me hods, 44,
287–304.
Kim, S. Y., Yap, M. J., & Goh, W. D. (2018). The ole o seman ic
anspa ency in isual wo d ecogni ion o compound wo ds:
A megas udy app oach. Beha io Resea ch Me hods, 51,
2722–2732.
Kupe man, V. (2015). Vi ual expe imen s in megas udies: A case
s udy o language and emo ion. Qua e ly Jou nal o Expe i-
men al Psychology, 68(8), 1693–1710.
Kupe man, V., S ad hagen-Gonzalez, H., & B ysbae , M. (2012).
Age-o -acquisi ion a ings o 30,000 English wo ds. Beha io
Resea ch Me hods, 44, 978–990.
Kuzne so a, A., B ockho , P. B., & Ch is ensen, R. H. B. (2017).
lme Tes Package: Tes s in Linea Mixed E ec s Models. Jou -
nal o S a is ical So wa e, 82(13), 1–26.
Lado -Weizman, Y., & Deu sch, A. (2022). The con ibu ion o con-
sonan s and owels o audi o y wo d ecogni ion is shaped by
language-speci ic p ope ies: E idence om Heb ew. Jou nal o
Expe imen al Psychology: Human Pe cep ion and Pe o mance,
48(5), 401.
Lallie , M., & Ca ei as, M. (2018). C oss-linguis ic ans e in bilin-
guals eading in wo alphabe ic o hog aphies: The g ain size
accommoda ion hypo hesis. Psychonomic Bulle in & Re iew,
25, 386–401.
Landis, J. R., & Koch, G. G. (1977). The measu emen o obse e
ag eemen o ca ego ical da a. Biome ics, 33(1), 159–174.
Le ne , I., A ms ong, B. C., & F os , R. (2014). Wha can we lea n
om lea ning models abou sensi i i y o le e -o de in isual
wo d ecogni ion? Jou nal o Memo y and Language, 77, 40–58.
Ma elli, M., & Amen a, S. (2018). A da abase o o hog aphy-
seman ics consis ency (OSC) es ima es o 15,017 English
wo ds. Beha io Resea ch Me hods, 50(4), 1482–1495.
Monsell, S., Doyle, M. C., & Hagga d, P. N. (1989). E ec s o e-
quency on isual wo d ecogni ion asks: Whe e a e hey? Jou -
nal o Expe imen al Psychology. Gene al, 118(1), 43–71.
Mo , B., & P io , A. (2020). Indi idual di e ences in L2 equency
e ec s in di e en sc ip bilinguals. In e na ional Jou nal o
Bilingualism, 24(4), 672–690.
Mo , B., & P io , A. (2021). F equency and p edic abili y e ec s in
i s and second language o di e en sc ip bilinguals. Jou nal
o Expe imen al Psychology: Lea ning, Memo y, and Cogni ion,
48(9), 1363.
Nema i, F., Wes bu y, C., Hollis, G., & Haghbin, H. (2022). The
Pe sian Lexicon P ojec : Minimized o hog aphic neighbou -
hood e ec s in a dense language. Jou nal o Psycholinguis ic
Resea ch, 51(5), 957–979.
New, B., Fe and, L., Pallie , C., & B ysbae , M. (2006). Reexamin-
ing he wo d leng h e ec in isual wo d ecogni ion: New e i-
dence om he English Lexicon P ojec . Psychonomic Bulle in
& Re iew, 13, 45–52.
Pee eman, R., & Con en , A. (1997). O hog aphic and phonological
neighbo hoods in naming: No all neighbo s a e equally in luen-
ial in o hog aphic space. Jou nal o Memo y and Language,
37(3), 382–410.
Pe ea, M., Abu Mallouh, R., & Ca ei as, M. (2014). A e oo le e s
compulso y o lexical access in Semi ic languages? The case
o masked o m-p iming in A abic. Cogni ion, 132, 491–500.
Pe ea, M., & Ca ei as, M. (2006a). Do ansposed-le e e ec s occu
ac oss lexeme bounda ies? Psychonomic Bulle in & Re iew,
13(3), 418–422.
Pe ea, M., & Ca ei as, M. (2006b). Do ansposed-le e simila i y
e ec s occu a a p elexical phonological le el? Qua e ly Jou -
nal o Expe imen al Psychology, 59(9), 1600–1613.
Pe ea, M., & Ca ei as, M. (2008). Do o ho ac ics and phonology
cons ain he ansposed-le e e ec ? Language and Cogni i e
P ocesses, 23(1), 69–92.
Pe ea, M., Mallouh, R. A., & Ca ei as, M. (2010). The sea ch o
an inpu -coding scheme: T ansposed-le e p iming in A abic.
Psychonomic Bulle in & Re iew, 17(3), 375–380.
Pe ea, M., & Lupke , S. J. (2004). Can CANISO ac i a e CASINO?
T ansposed-le e simila i y e ec s wi h nonadjacen le e posi-
ions. Jou nal o Memo y and Language, 51(2), 231–246.
Pe ea, M., & Lupke , S. J. (2003). Does jugde ac i a e COURT?
T ansposed-le e simila i y e ec s in masked associa i e p im-
ing. Memo y & Cogni ion, 31(6), 829–841.
Pe ea, M., Lupke , S. J., & Kinoshi a, S. (2003). T ansposed-le e
con usabili y e ec s in masked o m p iming.Masked p iming:
S a e o he a , 97–120.
Pei ce, J., G ay, J. R., Simpson, S., MacAskill, M., Höchenbe ge ,
R., Sogo, H., ... & Lindelø , J. K. (2019). PsychoPy2: Expe i-
men s in beha io made easy. Beha io Resea ch Me hods, 51,
195–203.
Pexman, P. M., Mu aki, E., Sidhu, D. M., Siakaluk, P. D., & Yap, M.
J. (2019). Quan i ying senso imo o expe ience: Body–objec
in e ac ion a ings o mo e han 9,000 English wo ds. Beha io
Resea ch Me hods, 51, 453–466.
Ras le, K., & B ysbae , M. (2006) Masked phonological p iming
e ec s in English: A e hey eal? Do hey ma e ? Cogni i e
Psychology, 53(2), 97–145.
Rau, A. K., Moll, K., Snowling, M. J., & Lande l, K. (2015). E ec s
o o hog aphic consis ency on eye mo emen beha io : Ge man
and English child en and adul s p ocess he same wo ds di e -
en ly. Jou nal o Expe imen al Child Psychology, 130, 92–105.
Ra id, D. D. (2011).Spelling mo phology: The psycholinguis ics o
Heb ew spelling(Vol. 3). Sp inge Science & Business Media.
Rueckl, J. G., Paz-Alonso, P. M., Mol ese, P. J., Kuo, W. J., Bick, A.,
F os , S. J., ... & F os , R. (2015). Uni e sal b ain signa u e o
p o icien eading: E idence om ou con as ing languages.
P oceedings o he Na ional Academy o Sciences, 112(50),
15510–15515.
Schmalz, X., Ma inus, E., Col hea , M., & Cas les, A. (2015). Ge ing
o he bo om o o hog aphic dep h. Psychonomic Bulle in &
Re iew, 22, 1614–1629.
Sch ö e , P., & Sch oede , S. (2017). The De elopmen al Lexicon P o-
jec : A beha io al da abase o in es iga e isual wo d ecogni ion
ac oss he li espan. Beha io Resea ch Me hods, 49, 2183–2203.
Schoonbae , S., & G ainge , J. (2004). Le e posi ion coding in
p in ed wo d pe cep ion: E ec s o epea ed and ansposed le -
e s. Language and Cogni i e P ocesses, 19(3), 333–367.
Seymou , P. H., A o, M., & E skine, J. M. (2003). Founda ion li e acy
acquisi ion in Eu opean o hog aphies. B i ish Jou nal o Psy-
chology, 94, 143–174.
Sha e, D. L. (1999). Phonological ecoding and o hog aphic lea ning:
A di ec es o he sel - eaching hypo hesis. Jou nal o Expe i-
men al Child Psychology, 72(2), 95–129.
8783Beha io Resea ch Me hods (2024) 56:8761–8783
Sha e, D. L., & Ba -On, A. (2018). Lea ning o ead a Semi ic abjad:
The iplex model o Heb ew eading de elopmen . Jou nal o
Lea ning Disabili ies, 51(5), 444–453.
Shim on, J. (2006).Reading Heb ew: The language and he psychology
o eading i . Rou ledge.
Seidenbe g, M. S. (2011). Reading in di e en w i ing sys ems: One
a chi ec u e, mul iple solu ions. In P. McCa dle, B. Mille , J.
R. Lee, & O. J. L. Tzeng (Eds.), Dyslexia ac oss languages:
O hog aphy and he b ain–gene–beha io link (pp. 146–168).
Paul H B ookes Publishing.
Shim on, J., & Si an, T. (1994). Reading p o iciency and o hog aphy:
E idence om Heb ew and English. Language Lea ning, 44,
5–27.
Siegelman, N., Kea ns, D. M., & Rueckl, J. G. (2020). Using in o ma-
ion- heo e ic measu es o cha ac e ize he s uc u e o he w i -
ing sys em: The case o o hog aphic-phonological egula i ies
in English. Beha io Resea ch Me hods, 52, 1292–1312.
Siegelman, N., Rueckl, J. G., Lo, J. C. M., Kea ns, D. M., Mo is,
R. D., & Comp on, D. L. (2022). Quan i ying he egula i ies
be ween o hog aphy and seman ics and hei impac on g oup-
and indi idual-le el beha io . Jou nal o Expe imen al Psychol-
ogy: Lea ning, Memo y, and Cogni ion, 48(6), 839.
Snell, J. (in p ess). PONG: A compu a ional model o isual wo d ec-
ogni ion h ough bi-hemisphe ic ac i a ion, Psychological Re iew.
Soa es, A. P., Lages, A., Sil a, A., Comesaña, M., Sousa, I., Pinhei o,
A. P., & Pe ea, M. (2019). Psycholinguis ic a iables in isual
wo d ecogni ion and p onuncia ion o Eu opean Po uguese
wo ds: A mega-s udy app oach. Language, Cogni ion and Neu-
oscience, 34(6), 689–719.
Sze, W. P., Ricka d Liow, S. J., & Yap, M. J. (2014). The Chinese Lexicon
P ojec : A eposi o y o lexical decision beha io al esponses o
2,500 Chinese cha ac e s. Beha io Resea ch Me hods, 46, 263–273.
Tsang, Y. K., Huang, J., Lui, M., Xue, M., Chan, Y. W. F., Wang, S.,
& Chen, H. C. (2018). MELD-SCH: A megas udy o lexical
decision in simpli ied Chinese. Beha io Resea ch Me hods, 50,
1763–1777.
Tse, C. S., Yap, M. J., Chan, Y. L., Sze, W. P., Shaoul, C., & Lin, D.
(2017). The Chinese Lexicon P ojec : A megas udy o lexical deci-
sion pe o mance o 25,000+ adi ional Chinese wo-cha ac e
compound wo ds. Beha io Resea ch Me hods, 49, 1503–1519.
Van Pa idon, J., & Thompson, B. (2021). subs2 ec: Wo d embeddings
om sub i les in 55 languages. Beha io Resea ch Me hods, 53,
629–655.
Velan, H., & F os , R. (2007). Camb idge Uni e si y e sus Heb ew Uni e -
si y: The impac o le e ansposi ion on eading English and Heb ew.
Psychonomic Bulle in & Re iew 2007 14:5, 14(5), 913–918.
Velan, H., & F os , R. (2011). Wo ds wi h and wi hou in e nal s uc-
u e: Wha de e mines he na u e o o hog aphic and mo pho-
logical p ocessing? Cogni ion, 118(2), 141–156.
Velan, H., Deu sch, A., & F os , R. (2013). The lexibili y o le e -
posi ion lexibili y: E idence om eye mo emen s in eading
Heb ew. Jou nal o Expe imen al Psychology: Human Pe cep-
ion and Pe o mance, 39(4), 1143.
Velan, H., F os , R., Deu sch, A., & Plau , D. C. (2005). The p ocess-
ing o oo mo phemes in Heb ew: Con as ing localis and dis-
ibu ed accoun s. Language and Cogni i e P ocesses, 20(1–2),
169–206.
Velan, H., & F os , R. (2009). ansposi ion e ec s a e no uni e sal:
The impac o ansposing le e s in Heb ew. Jou nal o Memo y
and Language, 61(3), 285–302.
Voe en, C. C. (2019). Using ‘buildme ’ o au oma ically ind & com-
pa e maximal (mixed) models. R Package Ve sion, 1(6), 1–7.
Weiss, Y., Ka zi , T., & Bi an, T. (2015). The e ec s o o hog aphic
anspa ency and amilia i y on eading Heb ew wo ds in adul s
wi h and wi hou dyslexia. Annals o Dyslexia, 65, 84–102.
Yap, M. J., Ricka d Liow, S. J., Jalil, S. B., & Faizal, S. S. B. (2010).
The Malay Lexicon P ojec : A da abase o lexical s a is ics o
9,592 wo ds. Beha io Resea ch Me hods, 42(4), 992–1003.
Ya koni, T., Balo a, D., & Yap, M. (2008). Mo ing beyond Col hea ’s
N: A new measu e o o hog aphic simila i y. Psychonomic Bul-
le in and Re iew, 15(5), 971–979.
Publishe 's No e Sp inge Na u e emains neu al wi h ega d o
ju isdic ional claims in published maps and ins i u ional a ilia ions.