SCIENTIFIC REPORTS
OPEN Enhancing reading performance
through action video games: the
role of visual attention span
Received: 16 August 2017 A.Antzaka*?, M. Lallier®4, 5. Meyer, J.
Accepted: 19 October 2017
Published online: 06 November 2017
4, M. Carreiras(** &5.Valdois**
Recent studies reported that Action Video Game- AVG training improves not only certain attentional
‘components, butalzo reading fluency in children with dyslexia. We aimed to investigate the shared
attentional components of AVG playing and reading, by studying whether the Visual Attention (VA)
‘span, acomponent of vicual attention that has previously been linked to both reading development and
dyslexia, i improved infrequent players of AVGs, Thirty-six French fluent adult readers, matched on
“chronological age and text reading proficiency, composed two groups: frequent AVG players and non.
players. Participants performed behavioural tacks measuring theVA span, and achallenging reading
‘task (reading of brieffy presented pseuco- words). AVG players performed betteron both tasks and
erformanceon these tasks was correlated. These results further support the transfer of the attentional
‘benefits of playing AVGs to reading, and indicate that the VA span could be acore component mediating
‘this transfer The correlation between VA span and pseudo-word reading ao supports the involvement
‘ofVAspaneven in adult reading. Future studies could combine VA span training with defining features
OfAVGs, inorcer to buikd anew generation of remediation software.
Action video games (AVGs) and books are very different and ye, the visual processes involved in playing AVG
and in reading a book could be closely linked, The former display complex scenes, sophisticated geometric ren
Sering and rapid moving objects, whereas the latter contain Black and white static print, usually ina singe alpha.
bet and font. However, Franceschini etal. showed a positive effect of AVG training on reading in dyslexic
children (se also Gori etal) A group of children played AVGs fr twelve hours. After training they signficanlly
{improved both thee visual attention and ther reading speed, without los of reading sccuracy. This observation
is puzzling. How could an offthe shelf AVG have such an impact on reading performance? Could AVGs be used
toimprove reading speed, without specifically targeting verbal material?
There ae several possible links between AVG playing and reading. AVGs ate defined by high speed events
and fast moving targets, spatial and temporal unpredictability, an emphasis onthe peripheral visual field, and
high motor, perceptual and cognitive loads" They are one of the most studied classes of video games because
oftheir positive elfect on various cognitive and perceptual processes. For example, AVG players are better in
contrast discrimination, probabilistic inference” and mental rotation tasks. The features of AVGs setup high
ftentional requirements to succeed atthe game, so we could assume that playing AVG specifically trains visual
tention. And indeed, it does: the causal link between AVG practice and improvement of visual attention has
been assessed in many studies” (for a review see ref). Visual attention covers large numberof dimensions, so
that AVG alfect many diferent tasks, To name a few, AVG players are better at visual earch", in enumeration
tasks and in multiple-object-tracking asks. They also ae less susceptible to crowding effects (Vora review see
ref") Interestingly individual improvement in reading performance was correlated with improvement in visual
attention inthe studies by Franceschini eal
Temporal and spatial visual atention is enhanced in AVG players. Their higher performance on tests of
visuo-attentiona skills (enumeration, multiple-object-tracking) was attributed to spatial attention improve:
‘ment This is in particular exemplified by thir ability to track more objects simultaneously than non player
(Other findings suggest postive effects onthe temporal dimension of visual attention, This isn particular exer
plified in the attentional blink paradigm in which a stream of lttersis briefly displayed, one after the other, and
*Basque Center on Cogntian, Brain and Language, 20008, San Sebastian, Spain ‘Departamento de LenguaVasca
y Comunicacién, UPVIEHU, 48340, Leioa, Spain. “Université Grenable-Alpes, LPNC, F-38040, Grenoble, France
cnits, LPNC UMRS205, F-38040, Grenoble, France. “ikerbasqve, Basque Foundation for Science, 46013, Bilbao,
‘Spain, Correspondence and requests for materiale should be addressed toAAA (email 3 antzaka@bcbl es)
cl REPORTS [714563 [DOILOLOSRISHIE96-017.5119.9 2ae
participants have to quickly shift from one letter feature (the colour ofa frst target) to another (the identity ofa
Second target). The momentary blink inattention, observed when the second target occurs in the few hundred
milliseconds following the coloured letter, is reduced in AVG players, suggesting faster temporal processing skills,
than in non player Iie thus well established that AVGs improve several facets of attention and that AVG players
benef from greater attentional resources
(On the other hand, many studies have shown that visuo-attentional skills are involved in normal" and
pathological reading’. Our previous findings, in particular, emphasized the link between the visual attention
(VA) span and reading performance’. VA span is defined asthe numberof distinct visual elements (Le. letters,
Jina reading context) that can be simultaneously processed in one fixation’. The size of the VA span reflects the
amount of attention capacity that is available for mul-lement processing" and is inked to superior parietal,
lobule (SPL) activation for pre-lexical orthographic processing’. In typical children/teenagers,alarger VA
span relates to faster and more accurate reading!" large VA span help readers to process larger orthographic
‘Unite"™". Individuals with alarge VA span can process most familiar words and within-unfamiliar- words longer
sub-lexical units (mult-letter graphemes or syllables) asa whole, which results in higher reading speed. Thus,
children with higher VA span process more eters at cach fixation and show faster text reading” The impact of
\VAspan on text reading further holds for adult readers The contribution of VA span to pseuda-svord reading is
independent of phonological skllsin typical readers” and a subset of dyslexic children exhibits a VA span deficit
and poor preudo-word reading despite good phonological skills” =",
‘Some authors argue that VA span impairment would be the consequence ofthe poor reading skill of dyslexic
readers but evidence from longitudinal and training studies speaks against sucha consequence link” The VA
span of pre-readers predicts ther future reading performance™ In pure visual attention span dyslexia, VA span
{taining has a significant impact on reading speed"™The fact thatthe VA span is invalved in non-verbal tasks
1nd non-verbal material"="" is further evidence against any interpretation thatthe VA span-teading relation-
ship is mediated by language or reading experience. Furthermore, the VA span deficit in developmental dyslexia
is not restricted to horizontal reading like displays, Children showing VA span limitations on horizontal let
{er strings are similarly impaired when using circular displays", They further shove a deficit in visual search
tasks for spatially distributed stimuli, which strongly speaks against specialization for reading and/or horizontal
array processing". Because the VA span is not restricted to verbal stimuli, and because it is not restricted to
‘mono: dimensional stim it could alo be used by players when processing visual simul presented by an AVC,
(Overall, AVG playing affects the spatial distribution of attention over the visual scene so tht a larger deploy
ment of attentional resources elps AVG players to process more vstal information simultaneously. On the other
hand, children with higher VA span have greater attentional resources which they deploy more widely to pro-
cest more letters simultancously within strings, leading in turn to faster and more accurate reading, When put
together, these results strongly suggest thatthe VA span might be a common component between playing AVG
and reading. AVG playing might enlarge the players VA span which would help them process larger mult-leter
‘units, and consequently, would improve their reading performance. The main aim ofthe current study is to test
this hypothesis. Iwo groups of AVG players and non-video game (NVG hereafter) players were recruited; they
were matched on chronological age and text reading proficiency. We assessed theit VA span abilities (Le, their
ability to process multiple elements simultaneously) using clasical partial and global report paradigms. During
partial (or global) report, a 6-consonant string is briefly presented tothe participant. One string position (or
‘none) is cued at each tial, and the participant has to name the cued letter only (or the whole sequence), Single lt
ter identification skills were further asested to ensure that differences in performance on the letter-string report
tasks were not explained by differences in single letter processing. Finally, we tested the participants ability to
ead briefly presented 6-ltte pseudo-words (PW) made of units of dlferent sizes. AVG players were expected to
show larger VA span and more accurate PW reading: they should also be more sensitive t larger PW stb-lesial
‘units than non-players
Results
Visual Attention SpanTasks. Global Report. Performance of the two groups of participants in global
report isllusrated in Fig and descriptive dates provided in Supplementary table
i the global report ek participants were asked to report ae many Kters a possible from 24 Geter strings
presented successively at the cente ofthe scren for 200 ms, The percentage of eters accurately identified by
Postion in the AVG (n= 19) and NVG (n= 17) groupe ieilustated in Fig. 1 A Type ill ANOVA was per-
Formed onthe original data with Group (AVG vs. NVG) athe between-sbjectfctr and Letter Position athe
‘within-subject factor (Positions 1-6) There wae a malin effet of Group (FC, > 2033)
anda ain effect of Letter Poston (FS, 170)~24.38, p=0.00, ty 0.68), The AVG group identiied more
letters accurately than the NVG group (M(SD)=91.41 (4.41) % vs. M(SD)~ 85.29 (6.4) %). The Group by Letter
Poston interaction was significant (HS, 170) =4.04, p= 0.002, r2—04). Post hoc comparisons onthe Group
by Letter Poston interaction indicated thatthe AVG and NVG¢ groups performed simriy onthe tree fest
postions ofthe string Positions 1, 2and alps > 0.19), but the AVG group dened more letters accurately on
Foston 4 (3-981 =2.,p~ 0.03), Poston 5 (313541457, p01) and Postion 6 (= 791,t=-287,
0008).
‘We reasoned that better parallel procesing in AVG players would result ina more homogencous spreading of
vinalattention across the whole string thus sling in more balanced identification of eters actos the hemi
Fes To est his hypothesis left-right hemifield comparison was performed while equaling or he two groups
vera performance onthe ek-The percent average accuracy of participant hele (postions 12 and : eft
Hemi) and right of nation (cro postions 43, and 6 righ hemifield) were compited, These ecores were
then divided bythe participants mean percent score, across ll positions. Thus, the sum of both scores fr cach
palicipant way 1 Therefor, the “Hemfica scores represented the relative peformance on each hemifield for
SCIENTIFIC REPORTS [714563 [DOLLO10391#2598-017-15119.9 2Sue
Proportion of Letters Identified
‘Number of Letters Identified (%)
Ti 3 @ 3 6 Tet
Right
Letter Position in the String Hemifield
Figure 1. Global eport task - Letter identification by position (1a) and lef-right hemifield relative
performance (1b) forthe AVG players (==) and non-players (+++). Error bars represent one standard error.
cach participant, egardes oftheir overall accuracy thus reflecting ther visual attention dsrbation strategy
ner the string Hemel eet on global report pevormance are lustrated in Fig Tb. Type ll ANOVA was
performed on the original data with Group (AVG vs, NVG) asthe beteen-sbjet factor and Hemifeld a the
Shin abject factor Av expected there wes no main elect of Group (F138) =0, py y=) det the
Inlied scores sed (se above). There was a main elect of Hemi (34) =2996,p “01,2 0.) and
2 Group by Hemieldinteracion (F, 38) 8.1, 19) The posthoe comparison onthe Groep
ty Hesmifeld interaction indicated that in both Croup, partkipants responded more acurately to ters p
Stated inthe et as compared to the ight hemifield (AVG 3008 ¢=754,p cODD1, NY
‘P= 0.001) Morcover the NVG paricpant hada restive performance cor above ta of the AVG participants
Ente left hemi (AVG-NVG P0006) bt below that ofthe AVG participants on the
Sight hemifield (AVG-NVG: 8 003,285, p 0.000), This sugests stronger let bis forthe NVG than the
NG particpants
Iemore widely spread across the letter string higher attentional resources would farther result in lesser
inter potion varaityon eter Kentication Ihe coreation between individual standard deviation scores
(tceting indivi inter postion variability n performance on the global report tak) and individ mean
Percent scores crow a pstons on he global report task was sigan buth vss he whole group of prt
77,9 001, and within each grovp of pati. 078 p 000
07% pe 000%, see Supplementary fg SI). Thisndictes tht smaller varsity elated
totter overall performancc onthe aks In particular the AVG group showed lower by-posion
performance than the NVG group (AVG: MOD) ~ 89 (3.29), NV MISD)~ 138115 98).
08
Partial Report. ln the partial report task, the participants were successively presented 72 6-letter strings briely
(200 ms) inthe centre ofthe screen. After the presentation of each string a relro-cuc indicated which letter they
should report from the presented string. The percentage of cued letters accurately identified by postion in the
AVG (n= 19) and NVG (n= 17) groups is illustrated in Fig. 2 and descriptive data is provided in Supplementary
Table $2 Type Ill ANOVA was performed onthe original data with Group (AVG vs. NVG) as a between-subject
‘within-subject factor (Positions 1-6). There was a main effect of Group (F(1,
34) =6.28, p=00017, ry!=0.16) and a main effect of Letter Position (F(3, 170) =8.70, p< 0.001, 1j2=0.55). AS
in global report, the AVG Group identified significantly more letters than the NVG Group (M(SD) =88.38 (5.95)
79.74 (13.66) %) but the Group by Letter Position interaction was not significant (F(S, 170) = L08,
27), meaning thatthe AVG players advantage was similar across the sx positions,
The correlation between individual standard deviation scores (reflecting individual interposition variability
in performance on the patil report task) with individual mean percent scores onthe partial report task was
significant both across the whole group of pa 6,75 = 0.73, p< 0.001), and within each grou
of participants (AVG( 0.62, p< 0.001; NVG(n= 17): %y= 0.77, p< 0.001, see Supplementat
Fig 82). Once again, as expected, based on the higher mean performance ofthe AVG as compared to the NVG
bility in performance (AVG: M(SD) = 10.85 (5.13), NVG:
r=-036),
SCIENTIFIC REPORTS [714563 [DOLLO10391#2598-017-15119.9 3Number of Letters Identified (%)
T 23 4 36
Letter Position in the String
Figure 2. Partial report tak - Letter identification by position for the AVG players (=) and non-plyers
(Gross), Error bars represent one standard error.
Control Tsk. "The singe eter (SL) dentifcation tsk was used as contol to ensure tat the AVG and NVG.
groups did not dller on single leter processing so that better performance on global and partial report in the
‘AVG proup cou be reliably interpreted as evidence of higher mull leltr processing sil Group comparison
revealed no diferenee between the AVG players and non-players(U=157-5,2-=-018, 088,703) on
their computed threshold (reflecting the shortest presentation Guration at which at leat 0% letters were ac~
rat identified) on thisask (AVG n=19, M(SD) = 36:58 (7.12). NVG:n-—=17, MSD) =37.00 (743). As further
{evidence forthe independence of performance in single and ml-leter prcessing,»vcighted sum of pelor-
‘ance onthe SL task score at 331ms* 5+ score at 50 ms "4+ score at 67 ms * 3+- score a 84m 2+ score at
TOI ms") was used wo correlate wih performance onthe global and paral report asks, The correlations were hat
significant ‘003, p=0.8) suggesting thatthe diferences between the wo
‘toups were not related to singe eter processing but are apparent only when mult letter processing is fequied
Pseudo-word Reading. The AVG and NVG players were asked to read aloud pseudo-words that were
briefly presented at the centre of the computer screen for 60 ms and followed by a mask. All pseudo-words
were 6-letters long but they varied in the numberof syllables either including three CV syllables (CVCVCV.
pseudo-words such as "siluve") or two syllables. In the latter case, the second syllable included along vocalic
sgrapheme (CVCVIgas in rigois") Half ofthe pseudo-words ofeach syllable length included an existing word
Corresponding to cither a CVCV word forthe 3 syllable-long pseudo-words (eg, rimode"), ora CVIg word for
{he 2-sllablelong pscudo-words (eg, gibois). The number of pseudo-words accurately named bythe AVG and
VG players is provided on Table |.
The generalized linar mixed elfects model wsed to analyse the accuracy of responses on each tril included
the fixed effects of Group (AVG vs. NVG), Word Presence (Word Present vs. Word Absent) and Grapheme Size
(presence ofa large grapheme or not) and their interactions. All factors were coded assum contrasts. The most
complex random effec structure that converged included random intercepts by subject and item and a random,
by subject slope for Grapheme Size, Information on the model i provided in Supplementary Table 83, The xclu-
sion of two outliers (one from the NVG and one from the AVG group) did not change the pattern of results and
significance soall participants were incladed inthe analysis.
Only the effect of Group vas significant with the AVG group naming 79.34% (SD = 14.87) pseudo-words
accurately against 68.44% (SD = 15.42) for the NVG group (3 = —0.37, 2= ~2.39, p=0.017). No effect of
pseudo -word structure/grapheme size (CVCVCV vs. CVCVg) or word presence within the pseudo-word was
found (ps > 0.19), suggesting that neither the AVG nor NVG group was sensitive tothe lexical units embedded
wwthin the pscudo-words
Correlations were computed between VA span scores (computed from performance on the global and partial
report reduced to the mean numberof letters accurately processed in each tral") and pscudo-word reading
accuracy The two outliers of the pecudo-word reading task were excluded and VA span scores were exponen-
tially transformed to improve ther distribution, The corelation was significant both within each group (AVG
(n= 18):r=041, p=0.047; NVG (n= 16):7=048, p= 0.047; Fig. 3) and acrossall participants (r(n=34) =0.60,
P<0000).
Inorder test the variance in preudlo-word reading accuracy explained by VA span skills a regression analysis
was performed including the data of both groups of participants. Once agai, the two outliers on pseudo- word
reading accuracy were removed and VA span scores were exponentially transformed, All the dependent variables
‘were centred. A first regression model included only the control variables chronological age nd the weighted
SCIENTIFIC REPORTS [714563 [DOLLO10391#2598-017-15119.9 4en 50) YOR a Dy
ava im=9)
ange aT [pasos fast a7
peonn SRT a7 i at a
ange eat [ps0-09 [arrose [ao
Table 1. Accuracy ofthe AVG and NVG group in pseudo-word (PW) reading
‘Vise Ateaton Span exp)
Figure 3. Correlations between composite measure of VA span and preudo-word reading accuracy fr the
ANG players (=) and non-players
score of single letter identification. ‘The second regression model included the control vat
scores, The models were compared using a chi-squared test and comparing the moltiple and adjusted r-squared
values ofthe two models. The model including only the control variables (R= 0.08, adjusted R?=0.02, F(2,
31)= 1.39, p=0.26) did not explain significant variance in pseudo-word naming accuracy while the model
including the control variables and VA span scores did explain significant variance (R?—0.37, adjusted R?=0.31,
LG, 30) =5:90, p=0.003). In line wth this result, the model including VA span scores had a significantly higher
goodness of il than the one including only the control variables (X=(1, n= 34) =0.14, p< 0.001), and in this
‘model the effect of VA span was significant (3=0.015, 1=3.71, p< 0.001). Importantly, when the categorical
factor Group (AVG vs. NAVG) vas added tothe first model withthe control variables, the model did explain
significant variance in pseudo-word naming accuracy (R° =0.29, adjusted R°=0.22, FS, 30) =4.02, p=0.016).
Nevertheless, the addition of the VA span to the model (R!—0.42, adjusted R —0.M, Fs, 28)
stlllead toa significant improvement (X(1, 0 =34) (0.011) The effect of VA span was also significant
inthis final model (3 =0.011,¢ (016), indicating that VA span sills explained addtional unique var-
face in pscudo-word naming accuracy after taking into account both the control variables and the group effect.
bles and VA span
Discussion
The present stad sed the hypothe that VA san woul bea rial etre xing the fle of AVG
ing on reading performance’, Previous evidence fora link between AVGs and reading performance was
reported in childs and developmental djlexa Wehere focused on two groups of youngadatc who wercelther
ANG players or non-players (NVG). Our first aim was to explore whether AVG players have higher VA span
skill than NVG players. Our second aim was toasters whether they performed better in a pscudo- word reading
task, thus providing first evidence thatthe effect of AVGs on reading extended to expert readers. In addition, we
explored the relationship between VA span and reading performance to verily whether fster and more accurate
reading in these groups of participants was related to larger VA span,
‘A first key finding of the current study is the larger VA span observed in AVG players compared to non-players
“The postive impact of AVG on visual processing and visual attention has been largely documented (Fora review
see ref") Individuals who play AVGs improve their visual sensitivity’ are less sensitive to visual interference (oF
crowding") and show enltnced temporal resolution in attentional blink tasks". However, none ofthese factors
‘an straightforwardly explain the higher performance of AVG players onthe VA span tasks, Higher sensitivity
to letter details might impact the participant’ ability to proces eter string, but the two groups of AVG players
and non- players were matched for leter identification skill. So, AVG players are more efictent in processing
leter-strings despite having single eter processing as fas asthe non players. Crowding can also alect eter,
string processing'"® but between-consonant spacing was increased in global and partial report t avoid crowd
ing elects, so that better performance of the AVG players on thes tasks could hardly be just the consequence of
lovter sensitivity to interference. Lastly, performance on the VA span tasks, where visual clement are simltanc-
‘ously displayed, has been shown to dissociate from processes invalving rapidly serially presented visual stimmai®
Instead, resource-based account of the VA span could explain the current link observed between AVG playing
and VA span, given thatthe connection between playing AVGe and benefits in attentional resources i already well,
SCIENTIFIC REPORTS [714563 [DOLLO10391#2598-017-15119.9 5ae
cstablished, Indeed, lgger VA spans on the global and the partial report tasks were shown orf the ll
Cation of greater attentional resources to multiple stimuli presented at once™ In ather words, participant with
Sager VA span can allocate more altentional resources to cach clement presented win the ting a parallel,
thus enhancing the numberof ters that ean be accurately identfed simbltancooay
To explore more in depth whether higher VA span in the AVG players resulted from the ability to allocate
‘more altentional sources across the letter sting, the esponse patter ofthe bwo groups inthe global and patil
report tasks was analyzed, In global report, sults revealed a weaker left basin the AVG group ax compared 19
{he NVG group, as expected following enhanced parallel processing. However, interpreting the esuls this way’
say be incoreect as performance was almost at cling forthe AVG group onthe let heme Interpretation of
performance in paral reports more straightforward Inthe absence of cling fect. AVG players showed an
{veal eter performance than NVG regardles othe letter target postion Th overall ndings strongly suggest
that AVG players can deploy greter attentional resources in parallel scrosstheleter string ofcitate cach eter
processing and dentiistion
Different attention distribution strategies depending on task-demands have previously been reported
Although parallel processing is involved in both global and partial report, a letward bis of covert tention
that relate tothe direction of eading i only observed in global report We postulate that the greater amount
of atlention resources avilable in the AVG players allowed them ta adopt a wider distribution oftheir VA span
resources, which facilitated the identification ofthe rightward letters whose identification is harder given the
Jnherent nature ofthe task In contrast to global report there is typically no positon bias in parial report since
the target eter position i indicated bya tetzo-cae displayed at the ost of the leter sting. Higher attentional
resources in AVG players thus yielded better processing ofthe trgetletere independently ftir positon sn the
string. Furthermore, n both global and partial report, higher performance related to lesser interposition varia-
bili etter ientifation, These overall ndings suggest that AVG players exhibit geste atentonal resources
than NVG players allowing them to allocates higher amount of attentional resources to cach poston of he leer
sing
‘Asccond evidence supporting link between AVG playing, VA span and reading, isthe group dilfereace
observed on preudo-word reading sis Iti worth noting that he wo groups were prot matched for text
reading proficiency. We reasoned that since the young adult AVG players ad learned to read prior to playing
AVG, they should have developed expert reading sills independently of AVG playing experience and should
have similar reading sills as the NVG players, Nonetheless, AVG playing during adolescence or ler may have
modified the VA span sil of AVG players, thus thir abit to process more lees simultaneously. However,
thie wasnt expected to impact the reading of el words which are procesed asa whole" snd reeruit minimal
attention once aquired" In contrast the amount of afenion resources avallable was expected to impact the
processing of unfamiliar words (or pscudo-words) not yet encoded in long-term memory Better pseudo-word
Fading performance was thus expectedin the AVG group. Ax preudo-words ae thought tobe serially processed
wile reading", we expected the AVG players to proces large sub-onts than the NVG participants, if more
$Mentional resources were avalable. Because participants in both groups were expet readers, we aministred
highly demanding preudo-word reading tsk where the stim were displayed for only 60 ms.
Importantly, the group of AVG players was fr better at reading the brely displayed preudo-words than the
[VG players but was not more scnstve othe pscudo-word sub-unit In both groupe, stil including along
sub-unit ether an embedded word or along lable) were nt read better than those that di nt inlude these
sub-unit This suggest thatthe pscudo-words were mainly processed as hoe without decomposing their
dnternal structure. Although this claim seems counterintuitive when considering the classical model of reading
(res), the multitzace memory model” predicts the parallel processing of pseudo-words when they a
bricly presented This model postulate thatthe deployment of attentional resources acros the entre eter
strings the ist step ofthe reading process, regardless the lexical stats ofthe tems (ods or pseudo- word)
Failimgto proces the letter sting aa whole makes the system svitch to an analytic mode, characterized by serial
processing of sub nits. The model postulates that switching tothe analytic from the global mode imposes some
oat on processing illustrated by longer procesing times, However, sich processing cox could hardly occ in 60
ts. The cureent findings thus suggest that AVG players are more prone to process pacudo-words accurately asa
‘whole than non-players.
as we explored whether AVG players witha large VA span demonstrated more accurate pseudo-word read
ing. Significant corcatons wer found, not onlin the AVG player group but alsa nthe nom players and in the
‘whole population Such VA spanreading relationship was previous) reported in youngadult expert readers as
vellasin aly and dysest cildren' ="
Overall ur ndings are consistent with Bavlir, Green & Scidenberg" proposal that AVG enhance visual
attention sills thos allowing the processing of larger mult-leter units in reading. More specifically, we pro-
oe that plying AVGs improve the ability to simultaneously process multiple elements in visual displays,
This benefit would increase VA span resources and significantly contribute to boosting reading performance
{hough the capacity of processing more letters n pralel within strings. This isnot to say that VA span isthe
tongue facet of visual attention tht i enhanced by AVGs andthe sale component that elates to reading sl
Franceschini etal have shown that playing AVGs improves focused visual altention sx well as working memory
and visual-to-audtory attentional shift. We jst claim here thatthe VA span is one of the critical features that
Contribute to explain why AVGs improve reading sil
‘Not that current evidence for parallel processing enhancement in AVG players would require controlling
snore strictly for other types visu spatilly demanding activities in the participants Indeed, some sport tain
dng and some diye situations, as ar deving may mor specically act parallel processing sil
‘We acknowledge thatthe current findings do not establish causal relationship between AVG practic, VA
span and reading Tlowever, ur results highlight the VA span as a rtcal feature improved by the practice of
SCIENTIFIC REPORTS [714563 [DOLLO10391#2598-017-15119.9ae
AVG that trigger faster and more efficient reading, Training studies are needed to establish causality in showing
that initially non players trained with AVGs develop a larger VA span and better reading skill than children.
trained with non-AVGs. We remain however very confident thatthe relationship is causal. Firt, previous stud-
jes on AVGs have shown that the results of training stadies are nearly identical to the reslts obtained through
comparison of groups of previously AVG players and NVGG players. In particular, there is no evidence that AVG,
players are individuals who had improved visual attention skills prior to becoming players. Second, both VA span.
find reading skills improve in dyslexic children when trained to simultancously process briefly presented visual
‘multiclements, be they alphanumeric or not, thus using training programs that share some common features
with AVGs The current findings pave the way fora new generation of training programs that should be devel:
‘oped to improve reading acquisition and remediate developmental dyslexia focusing on VA span training while
‘dopting the defining features of AVGs.
Methods
Participants. A total of 38 French right-handed adults (18-45 years old) with normal or corrected-to-nor-
sal vision, were recruited through advertising atthe Grenoble Alpes University Campus and the RISC (Relais
information sur les sciences de la cognition) mailing list. Recruitment targeted two groups of participants: a
group that didnot play action video games, named the NVG group, and a group of action video game players,
called the AVG group. To be included in the latter group, participants had to have played action video games
regularly (at least Shours 2 week) during the six months prior to the study. Information regarding language
background, handedness and experience playing action video games was acquired based on a questionnaire (see
Supplementary Methods S1). The questionnaire was adapted from previous studies! Participants reading pro-
ficiency was assessed using the text reading task ofthe Eclat- 16+ battery”. The test equired reading a text a=
quickly and as accurately as possible for !minute. The numberof words correctly read (total words read-errors)
pet minute was calculated fr each participant.
A total of 36, out ofthe total 38 young adults who were recruited, complied withthe recruitment criteria and
participated in the study. Nineteen participants (14 males) who reported playing AVG regularly (mean hours
played per month — 69.94, SD ~37.2.range: 20-173hours, the hour reported by two participants were removed
from these descriptive statistics since they were outliers, eporting 360 and 291 hours played per month") were
included in the AVG geoup (mean age = 20.89, SD =2.66, range: 18-26)-Seventeen participants (7 males fllinto
the NVG category (mean age = 20.76, SD =2.84, range: 18-28) since they reported either no video game practice