Hedenius, M., Persson, J., Alm, P.A., Ullman, M.T., Howard, J.H., Howard, D.V., & Jennische, M.

12 Pages·2013·0.77 MB·English

by  Martina Hedenius

ResearchinDevelopmentalDisabilities34(2013)3924–3935
ContentslistsavailableatScienceDirect
Research in Developmental Disabilities
Impaired implicit sequence learning in children with
developmental dyslexia
Martina Hedeniusa,*, Jonas Perssonb, Per A. Alma, Michael T. Ullmanc,
James H. Howard Jr.d,e,f, Darlene V. Howardd, Margareta Jennischea
aUnitforSpeechandLanguagePathology,DepartmentofNeuroscience,UniversityofUppsala,P.O.Box256,SE-75105Uppsala,Sweden
bAgingResearchCenter,KarolinskaInstituteandStockholmUniversity,Ga¨vlegatan16,SE-11330Stockholm,Sweden
cBrainandLanguageLab,DepartmentofNeuroscience,GeorgetownUniversity,37thandOStreets,N.W.,Washington,DC20057,USA
dDepartmentofPsychology,GeorgetownUniversity,37thandOStreets,N.W.,Washington,DC20057,USA
eDepartmentofNeurology,GeorgetownUniversity,Washington,DC,USA
fDepartmentofPsychology,CatholicUniversityofAmerica,620MichiganAvenue,N.E.,Washington,DC20064,USA
ARTI CLE INFO ABS TRA CT
Articlehistory: It has been proposed that an impairment of procedural memory underlies a range of
Received18June2013 linguistic,cognitiveand motor impairments observedindevelopmental dyslexia (DD).
Receivedinrevisedform15August2013 However,studiesdesignedtotestthishypothesisusingtheimplicitsequencelearning
Accepted15August2013 paradigmhaveyieldedinconsistentresults.Afundamentalaspectofprocedurallearningis
Availableonline thatittakesplaceoveranextendedtime-periodthatmaybedividedintodistinctstages
basedonbothbehaviouralcharacteristicsandneuralcorrelatesofperformance.Yet,no
Keywords: study of implicit sequence learning in children with DD has included learning stages
Developmentaldyslexia beyondasinglepracticesession.Thepresentstudywasdesignedtofillthisimportantgap
Proceduralmemory
byextendingtheinvestigationtoincludetheeffectsofovernightconsolidationaswellas
Implicitsequencelearning
those of further practice on a subsequent day. The results suggest that the most
pronounced procedural learning impairment in DD may emerge only after extended
practice,inlearningstagesbeyondasinglepracticesession.
(cid:2)2013ElsevierLtd.Allrightsreserved.
1. Introduction
Developmental dyslexia (DD) is characterized by unexpected difficulties with reading, in the context of typical
educationalopportunitiesand intactintellectual and sensoryabilities(Lyon,Shaywitz,& Shaywitz,2003).The disorder,
whichhasastronggeneticcomponent(Hensler,Schatschneider,Taylor,&Wagner,2010),hasbeenestimatedtoaffectabout
5–12%ofchildren(Shaywitz,Shaywitz,Fletcher,&Escobar,1990).ChildrenwithDDhavedifficultieswithwrittenword
recognitionandphonologicaldecoding(usingletter-soundmappingknowledgetodecodenovelwords),whichiswidely
believed to result from underlying phonological impairments (Snowling, 2000; Stanovich, 1988; Vellutino, Fletcher,
Snowling,&Scanlon,2004).
* Correspondingauthorat:UnitforSpeechandLanguagePathology,DepartmentofNeuroscience,Box593,UppsalaUniversity,75124Uppsala,Sweden.
Tel.:[phone];fax:[phone].
E-mailaddresses:[email protected](M.Hedenius),[email protected](J.Persson),
[email protected](P.A.Alm),[email protected](M.T.Ullman),[email protected](J.H.HowardJr.),[email protected](D.V.Howard),
[email protected](M.Jennische).
0891-4222/$–seefrontmatter(cid:2)2013ElsevierLtd.Allrightsreserved.
http://dx.doi.org/10.1016/j.ridd.2013.08.014
M.Hedeniusetal./ResearchinDevelopmentalDisabilities34(2013)3924–3935 3925
However,whilephonologicalimpairmentsareindeedfoundinanoverwhelmingmajorityofstudiesofDD(Ramus&
Ahissar, 2012), otherimpairments,whichare noteasily explainedby a specific phonologicaldeficit,arealso commonly
reported.Theseincludeimpairmentsofmotorfunctions(Nicolson,Fawcett,&Dean,2001),workingmemory(Smith-Spark&
Fisk,2007;Swanson,Xinhua,&Jerman,2009),executivefunctions(Brosnanetal.,2002),oculomotorandvisuoperceptual
functions(Quercia,Feiss,&Michel,2013),implicitsequencelearning(Howard,Howard,Japikse,&Eden,2006;Jimenez-
Fernandez, Vaquero, Jimenez, & Defior, 2011; Vicari, Marotta, Menghiai, Molinari, & Petrosini, 2003), artificial grammar
learning(Pavlidou,Williams,&Kelly,2009)aswellasproblemswithotheraspectsoflanguagethatappeartobeprimaryin
nature(i.e.notonlyaconsequenceofimpairedreading;Lyytinenetal.,2004;Snowling,Gallagher,&Frith,2003;Wimmer&
Schurz,2010).
Thispatternofwide-rangingimpairmentshasencouragedattemptstoprovideaunitaryexplanationforDDthatmay
account for both the phonological and non-phonological deficits in the form of a more general underlying deficit. Such
theoretical accounts include the proposals that DD is the result of impaired temporal perception (Tallal, 1980), of an
abnormaldevelopmentofthebrain’smagnocellularsystems(Stein,2001),ofadeficitinattentionalmechanisms(Hari&
Renvall,2001),oringeneralprocessingspeed(Wolf&Bowers,1999).Oneinfluentialtheoreticalview,whichisthefocusof
thepresentstudy,positsthattheunderlyingdeficitinDDiscausedbyadysfunctionintheproceduralmemorysystem,
specificallytothecortico-cerebellar(Nicolson,Fawcett,Brookes,&Needle,2010;Nicolsonetal.,2001)and/ortothecortical–
striatal(Ullman,2004)circuitsinthebrain.
Theproceduralmemorysystemunderliesthenon-declarative/implicitacquisition,consolidationandprocessingofskills
andhabits(Gabrieli,1998;Henke,2010;Squire&Zola,1996;Willingham,Salidis,&Gabrieli,2002).Thesystemreliesona
networkofbrainstructuresinwhichthecortico-striatalandcortical–cerebellarcircuitsplaycrucial,andlargelyoverlapping,
roles(forareview,seeDoyonandcolleagues,2009).
Althoughpreviouslyconsideredtobeimportantmainlyformotorfunctions(suchaslearninghowtorideabicycle),itis
becomingincreasinglyclearthatthissystemalsounderliesarangeofperceptual,cognitive,andlinguisticskills.Alarge
literaturesuggeststhattheproceduralmemorysystemplaysacrucialroleinthelearningandcomputationofsequences
(Aldridge&Berridge,1998;Knowlton,Mangels,&Squire,1996;Poldrack,Prabhakaran,Seger,&Gabrieli,1999;Saint-Cyr,
Taylor, & Lang, 1988; Willingham et al., 2002). This system also appears to be important for other functions, including
statisticallearning(McNealy,Mazziotta,&Dapretto,2010;Reeder,Newport,&Aslin,2013;Saffran,Aslin,&Newport,1996),
probabilisticclassificationlearning(Poldracketal.,2001;Poldrack&Rodriguez,2004),andtaskstappingcomplexlearned
motorskills(Ullman&Pierpont,2005).Accumulatingevidenceindicatesthatproceduralmemoryalsounderliesthelearning
anduseofrule-governedaspectsofgrammar,acrosssyntax,morphologyandphonology(Conway,Bauernschmidt,Huang,&
Pisoni, 2010; Conway & Pisoni, 2008; Dominey, Hoen, Blanc, & Lelekov-Boissard, 2003; Karuza et al., 2013; Teichmann,
Dupoux,Kouider,&Bachoud-Levi,2006;Ullman,2001,2004;Ullmanetal.,1997;Ullman&Pierpont,2005).
Anextensivelyusedtaskforthestudyofnon-languageproceduralmemory,specificallyimplicitsequencelearning,isthe
serialreactiontime(SRT)task(Nissen&Bullemer,1987).Inthistask,participantsaretypicallyshownfourboxesorcircles
arrangedhorizontallyacrossacomputerscreen.Wheneverastimulusappearsinoneofthefourpositions,subjectsareto
pressoneoffourcorrespondingresponsekeysasquicklyandaccuratelyaspossible.Intheimplicitversionofthistask,
participantsarenottoldthatthestimuliarepresentedaccordingtoafixedsequence(asopposedtotheexplicitversion,in
which the sequential pattern is verbalized and memorized prior to practice). Sequence learning is operationalized as
improvementsintheaccuracyand/orreactiontimesofresponsestothesequence,ascomparedtorandomlyordereditems
introducedasacontrolconditionattheendofpractice.Whenadministeredasanimplicittask,learningintheSRTtask
appearstobelargely,thoughnotcompletely,incidentalandnon-conscious(Howard&Howard,1992;Willingham,Nissen,&
Bullemer,1989).
AfundamentalaspectofsequencelearningintheSRTtask,andofprocedurallearningmoregenerally,isthatittakesplace
overanextendedtime-period.Thisperiodmaybedividedintodistinctstagesonthebasisofbothbehaviouralcharacteristics
andneuralcorrelatesofperformance(Debasetal.,2010;Hauptmann,Reinhart,Brandt,&Karni,2005;Korman,Raz,Flash,&
Karni,2003;Orbanetal.,2010;Robertson,Pascual-Leone,&Miall,2004).
Typically,aninitialfastacquisitionstage,characterizedbyarapidimprovementinperformance(asevidencedbya
decrease in both response speed and errors), is followed by a gradual decrease in the learning rate and a trend
towardsanasymptote(Hauptmannetal.,2005;Kormanetal.,2003).Thisasymptoticshapeofthelearningcurvehas
been suggested to reflect a saturation of learning that appears to be necessary for consolidation processes to occur
normally(Hauptmann&Karni,2002;Hauptmannetal.,2005;Karnietal.,1998).Consolidationreferstotheprocess
by which an initially labile memory trace becomes more robust and resistant to interference (Doyon et al., 2009;
Robertsonetal.,2004).Sometimesconsolidationinvolvesanactualincreaseinperformance,withoutfurtherpractice,
a phenomenon referred to as off-line learning (Hauptmann et al., 2005; Nemeth et al., 2010; Song, 2009; Song,
Howard,&Howard,2007).Theendpointofprocedurallearningisautomaticityofthelearnedbehaviour.Whenaskill
isautomatizeditcanbeperformedeffortlesslyevenwhenattentionisdirectedelsewhere(asindualtasksituations;
Seger & Spiering, 2011).
BrainimagingstudiessuggestthatimplicitsequencelearningintheSRTtaskdependslargelyonproceduralmemory
brainstructures,inparticularthestriatum,cerebellum,associatedmotorcorticalregions,aswellasportionsofprefrontal
andparietalcortex(Grafton,Hazeltine,&Ivry,1995;Peigneuxetal.,2000;Rauchetal.,1997;forareview,seeDoyonand
colleagues, 2009). In addition, recent studies have highlighteda role for the medial temporal lobe in sequence learning
3926 M.Hedeniusetal./ResearchinDevelopmentalDisabilities34(2013)3924–3935
(Albouyetal.,2008;Gheysen,VanOpstal,Roggeman,VanWaelvelde,&Fias,2011;Schendan,Searl,Melrose,&Stern,2003;
Simon,Vaidya,Howard,&Howard,2012)underbothexplicitandimplicitconditions.
Importantly,theneuralcorrelatesofsequencelearninghavebeenshowntobemodulatedbytheamountofpracticewith
thetask;whereasearlylearningischaracterizedbyaratherwidespreadactivationpatternthatincludesthemedialtemporal
lobe,prefrontalcortex,striatumandcerebellum,studiesincludingextendedpractice,andlaterlearningstages,suggestan
increasingly important role for the striatum in learning stages beyond the fast acquisition stage (Doyon et al., 2009;
Rieckmann&Backman,2009;Rieckmann,Fischer,&Backman,2010;Simonetal.,2012).
StudiesinvestigatingimplicitsequencelearninginDDusingvariantsoftheSRTtaskhavenotyieldedconsistentresults.
Althoughthemajorityofpublishedstudieshavefoundasequencelearningdeficitinadults(Du&Kelly,2013;Howardetal.,
2006;Menghini,Hagberg,Caltagirone,Petrosini,&Vicaria,2006;Stoodley,Harrison,&Stein,2006)andchildren(Jimenez-
Fernandezetal.,2011;Vicarietal.,2003,2005)withDD,intactperformancehasalsobeenreported(Deroostetal.,2010;
Kelly,Griffiths,&Frith,2002;Russeler,Gerth,&Munte,2006).
Ithasbeenproposedthatthetypeofsequenceusedaswellasthetimeintervalexamined,couldaccountforsomeofthis
inconsistency (Du & Kelly, 2013; Orban, Lungu, & Doyon, 2008). Nevertheless, previous studies have focused almost
exclusivelyonlearninginasinglepracticesession,andthefastacquisitionstage,whileneglectingovernightconsolidation
andlaterlearning.ThisnarrowexperimentalfocusistroublesomesincetheeffectofDDonperformanceinlaterlearning
stagesshouldbeoftheoreticalaswellasclinicalandpedagogicalinterest.
ThepresentstudyextendedtheexaminationofprocedurallearninginchildrenwithDDtoincludenotonlyaninitial
practicesession,butalsoovernightconsolidationandfurtherpracticeonasubsequentday.Toourknowledge,ourstudyis
thefirsttoinvestigatetheseimportant,andpotentiallytheoreticallyinformative,aspectsofproceduralmemoryfunctionsin
childrenwithDD.Inaddition,weusedavariantoftheSRTtaskthathasnotpreviouslybeenusedwithchildrenwithDD,the
alternatingserialreactiontime(ASRT)task(Howardetal.,2004;Howard&Howard,1997).
TheASRTtaskhasanimportantadvantageovertheSRTtasksusedinpreviousstudies;intheASRTtaskrandomtrialsare
alternatingwithtrialsfollowingthefixedsequencethroughoutthetask,e.g.1-r-2-r-4-r-3(wherethenumberscorrespondto
locationsonthescreenandrstandsforrandomlocationsamongthefourpositions).Thealternatingstructureinthisdesign
allows for an assessment of sequence specific learning separate from general motor skill learning (i.e. the overall
improvementinresponsespeedthatisduetopracticewiththetask).Moreover,itispossibletoexaminesequencelearning
continuouslythroughoutthetask,ratherthanonlyatasinglepointattheend ofthetask.Thepresentstudymaythus
providenewinformationaboutthenatureofapotentialprocedurallearningdeficitinDD.
2. Method
2.1. Ethicsstatement
ThestudywasapprovedbytheethicalreviewboardinthecityofUppsala.Allparentsorguardiansprovidedinformed
writtenconsent;childrenprovidedinformedwrittenassentandreceivedacinematicketfortheirparticipation.
2.2. Participants
Twelvechildrenwithdevelopmentaldyslexia(DD)and17typicallydeveloping(TD)controlchildrenparticipatedinthe
study.Thegroupswerematchedforsex,ageandhandedness(Table1).
ChildrenwithDDwererecruitedviaspeech–languagepathologyclinicsinthecitiesofStockholm,Uppsala,Ga¨vleand
Va¨stera˚s,inSweden.AllchildrenwithDDhadbeenindependentlytestedanddiagnosedwithdyslexiabyacertifiedspeech–
language pathologist within 1.5 years prior to participation. The TD group consisted of a subset of children who were
Table1
Participantdemographicsandcognitivecharacteristics.
Variable DD(n=12) TD(n=17) Comparison
Mean SD Mean SD t p
Ageinyears 11.0 0.71 11.1 0.68 0.388 .701
Sex(f/m) 5/7 5/12 x2=0.47 .494
Handedness 85.1 16.4 92.2 10.2 1.42 .167
PIQ 87.9 12.1 97.1 15.0 1.74 .093
Phonologicaldecoding 1.75 0.87 5.24 1.15 8.87 <.001
Orthographicreading 1.92 1.16 5.76 1.15 8.84 <.001
NWR 106 5.2 111 5.4 2.29 .030
TROG 57.9 30.1 75.6 18.4 1.96 .060
PPVT 152 16.4 160 13.5 1.47 .154
Abbreviations:PIQ,PerformanceIQ(Raven,1998);NWR,Nonwordrepetition(Wassetal.,2008);TROG,TestforReceptionofGrammar(Bishop,1982);PPVT,
PeabodyPictureVocabularyScale–ThirdEdition(Dunn&Dunn,1997).Note:HandednessscoresarebasedontheEdinburghHandednessInventory
(Oldfield,1971).
M.Hedeniusetal./ResearchinDevelopmentalDisabilities34(2013)3924–3935 3927
recruitedfromschoolsinandaroundthecitiesofStockholmandUppsalaaspartofalargerstudyonmemoryandlanguagein
typicallydevelopingchildren.AllchildreninthestudywerereportedbytheirparentstobemonolingualSwedish-speaking,
tohavenormal(orcorrectedtonormal)visionandhearing,andtohavenoknowncognitiveormotorimpairment,exceptfor
readingproblemsintheDDgroup.
Participants were given a set of behavioural tests in order to characterize their vocabulary (Dunn & Dunn, 1997),
grammatical comprehension (Bishop, 1982; Holmberg & Lunda¨lv, 2002), verbal working memory (nonword repetition;
Wassetal.,2008),orthographicreading,phonologicaldecoding,andperformanceIQ(PIQ;Raven,1998)Asexpected,there
weresignificantgroupdifferencesforthetworeadingmeasuresaswellasforverbalworkingmemory.Bycontrast,thetwo
groups did not differ significantly in vocabulary, grammatical comprehension or PIQ, although the differences in both
grammaticalcomprehensionandPIQapproachedsignificance(Table1).
PerformanceIQwasincludedfordescriptiveratherthanexclusionarypurposesinthisstudysincewewantedourDD
sampletoreflectthepopulationofchildrenwhoareclinicallyidentifiedanddiagnosedwithDDinSweden.Inlinewith
evidencesuggestingthatthereisaweak,ifany,relationshipbetweenPIQandthereadingproblemscharacteristicofdyslexia
(Ferrer,Shaywitz,Holahan,Marchione,&Shaywitz,2010;Rispens,vanYperen,&vanDuijn,1991),PIQisnotusedasan
exclusionaryfactorforaclinicaldiagnosisofDDinSweden.ThePIQrangeintheDDgroupwas70–115andtheTDrangewas
80–140.
ThetworeadingtestswerepaperandpencilSwedishadaptations(Olofsson,2003)ofthecomputerizedphonological
decodingandorthographicreadingtasksdevelopedbyOlson,Forsberg,Wise,andRack(1994).Inthephonologicaldecoding
test,thetaskwastodecide,andunderlinewithapencil,whichoneofthreeorfourpseudo-wordswasapseudo-homophone
of a real word. (i.e. ‘‘sounds’’ like a real word). The score corresponds to the number of correctly identified
pseudohomophones within 2min, with a maximum score of 80. In the orthographic reading test, participants were
askedtounderlinethetruewordintrueword-pseudohomophonepairs.Becausethephonologicalcodesforthepairswere
identical,thewordanditspseudo-homophonewouldbepronouncedthesameinSwedish.Thus,inordertomakeacorrect
responsesubjectshadtouseword-specificorthographicknowledge.Thescorewasthenumberofcorrectlychosenwordsin
2min,withamaximumscoreof120.
AllTDchildrenhadstaninescores(cid:2)4outof9onbothreadingtests(correspondingtoperformanceatorabove–0.75SD).
AllchildrenintheDDgrouphadstaninescoresof(cid:3)3onbothtests,exceptforonechildwhohadastaninescoreof5onthe
orthographic reading test. The performance of this child is consistent with previous evidence suggesting that the
phonological decoding problems characteristic of DD can sometimes occur together with intact or even superior
orthographicskills(Siegel,Share,&Geva,1995).
2.3. Stimuliandprocedure
SequencelearningwastestedwithaversionoftheAlternatingSerialReactionTime(ASRT)task(Howardetal.,2004;
Howard & Howard, 1997). As with classic SRT tasks, a sequence of visual stimuli appeared in one of four horizontally
arrangedlocations(indicatedbyopencircles)onthecomputerscreen.Thestimuluswasapictureofadog.Subjectswere
askedtopressoneoffourhorizontallyarrangedbuttonswheneverthestimulusappearedinthecorrespondinglocationon
thescreen.Specifically,theywereaskedto‘‘catchthedog’’asquicklyandaccuratelyaspossiblebypressingthebutton
correspondingtothecircleinwhichthedogappeared(Hedeniusetal.,2011).Thesequenceregularitywasnotmentioned.
Subjectswereinstructedtousethemiddleandindexfingersofbothhands,andresponseswerecollectedwithaPSTSerial
ResponseBox.Stimuliwerepresented(onaLCDscreen),andresponsetimeswereacquired,withE-Primeversion1.2.
Targetlocationsweredeterminedbyarepeatingeight-elementstructureinwhichfixedandrandomlocationsalternated.
Allparticipantsreceivedthesame8-itemsequencepattern,1r2r4r3r.Thatis,thedogappearedinposition1(theleft-most
circle),thenrandomlyinanyofthefourcircles,theninposition2,andsoon.Onrandomtrials,theeventsweresampledfrom
auniformdistributionsuchthatthefourlocationswereequallylikely.
Eachstimuluspresentationandresponseconstitutedonetrial.Trialswereorganizedintoblocksof85trials:5warm-up
randomtrials (notincludedinanalyses)followedby 10repetitionsof the8-itemsequence(Howardet al.,2004).Short
breakswereofferedbetweenblocks.Feedbackwasgivenattheendofeachblock,toguidesubjectstoanaccuracylevelof
about92%.Allinstructionsandfeedbackweredisplayedvisuallyonthescreenaswellasreadaloudtotheparticipants.The
taskwasself-paced,suchthatthecorrectbuttonhadtobepressedbeforeanewstimuluswouldappearonthescreen.
However, the experimenter controlled the beginning of each new block by clicking the mouse. Reaction time (RT) was
measuredfromtargetonsettothefirstresponse.Thenextstimulusfollowedthecorrectresponseafterafixed120msdelay.
Testingwascarriedoutovertwoseparatesessionsonconsecutivedays.Onday1,subjectscompleted20blocks(1700
trials and 200 repetitions of the 8-item sequence). Twenty-four hours later they completed 5 blocks (425 trials and 50
repetitionsofthesequence).
2.4. Statisticalanalysis
FollowingpreviousASRTstudies,trialsthatconstitutedthefinaltrialin‘‘trills’’(e.g.121)and‘‘repetitions’’(e.g.111)were
excluded,astheyhavebeenshowntobeassociatedwithpre-existingresponsetendencies(fordetails,seeHowardetal.,
2004).ForRTs,medianvalueswerecalculatedseparatelyforpatternandrandomtrialsforeachblockandeachsubject.Next,
3928 M.Hedeniusetal./ResearchinDevelopmentalDisabilities34(2013)3924–3935
thesemedianvalueswereaveragedacross10consecutiveblocksinsession1,andacrossthe5blocksinsession2,inorderto
obtainmeanvaluesforpatternandrandomtrials,againforeachsubject,foranearlylearningstage(stage1:blocks1–10),an
intermediatelearningstage(stage2:blocks11–20)andalatelearningstage,whichoccurredafteranovernightinterval
(stage3:blocks21–25).Asimilardatareductionprocedurewasperformedonaccuracy.
IntheASRTtask,sequencespecificlearningistypicallyexpressedasanincreasingdifferencebetweentrialsthatfollowthe
sequentialpatternandtrialsthatarerandom.ForRTs,sequencelearningisseenasincreasinglyfasterresponsestopattern
comparedtorandomtrials.Foraccuracy,sequencelearningistypicallyexpressedasaprogressivedecreaseintheaccuracy
ofresponsestorandomtrialswhereasaccuracyforpatterntrialstendstoremainstablethroughoutthetask.Generalskill
learning, on the other hand, is seen in the form of the sequence independent (across both pattern and random trials)
reductioninresponsetimeasafunctionofpractice.
Groupdifferencesingeneralskilllearningand/orsequencespecificlearningwereexaminedwith2(group:DDvsTD)(cid:4)2
(trial type: pattern vs random)(cid:4)3 (learning stage: 1–3) mixed design ANOVAs, for RT and accuracy, with group as a
between-subjectvariableandtrialtypeandlearningstageaswithin-subject(repeatedmeasures)variables.
3. Results
3.1. Reactiontimeanalyses
ThetwogroupsdidnotdifferwithrespecttooverallRT(RTinmilliseconds(ms):DDmean=450,SD=93;TDmean=433,
SD=59,maineffectofgroup:F(1,27)<1).ConsistentwithFig.1,generalskilllearningwasseenasasignificantmaineffect
oflearningstage(F(1,54)=84.9,p<.0001,h 2=.759),inwhichresponsesforbothpatternandrandomtrialsbecamefaster
p
withpractice.Thelackofanygroup(cid:4)learningstageinteraction(F(2,54)<1)suggeststhatthiseffectwassimilarinthetwo
groups.
EvidenceforsequencespecificlearningwasfoundinthatRTsweresignificantlyfasterforpattern(mean=438,SD=73)
comparedtorandom(mean=442,SD=75)trialsacrossthetaskasawhole(maineffectoftrialtype:F(1,27)=6.56,p=.016,
h 2=.195).Thiseffectwasstablethroughoutthetaskanddidnotincreasesignificantlywithpractice(non-significanttrial-
p
type(cid:4)learningstageinteraction: F(2,54)<1).Ofparticular interesthere,thetwo groupsdid notdifferwithrespectto
sequencelearningeffectsonRT(group(cid:4)trial-typeinteraction:F(1,27)<1;group(cid:4)trial-type(cid:4)learningstageinteraction:
F(2,54)=1.51,p=.230,h 2=.053).
p
Insum,thegroup(cid:4)trialtype(cid:4)learningstageANOVAwithRTasthedependentvariablerevealedsignificantgeneralskill
learningthatcontinuedtodevelopthroughoutthetask.Evidenceforsequencespecificlearningwassomewhatweakerasit
wasobservedonlyasamaineffectacrossthetaskaswhole,andtherewasnoevidenceforanincreaseinsensitivitytothe
sequenceregularitywithpractice.Notably,theRTanalysesdidnotrevealanygroupdifferencesinoverallRT,generalskill
learning,orsequencespecificlearning.
[(Fig._1)TD$FIG]
Fig.1.Reactiontimesinmillisecondsforpatternandrandomtrialsasafunctionoflearningstageforthedevelopmentaldyslexia(DD)andtypically
developing(TD)groups.Thefiguredisplaysmeansandstandarderrorsforthethreelearningstages.
[(Fig._2)TD$FIG]
M.Hedeniusetal./ResearchinDevelopmentalDisabilities34(2013)3924–3935 3929
Fig.2.(A)Accuracyinpercentforpatternandrandomtrialsasafunctionoflearningstageforthedevelopmentaldyslexia(DD)andtypicallydeveloping(TD)
groups.(B)Accuracydifferencebetweenpatternandrandomtrialsasafunctionoflearningstageforthedevelopmentaldyslexia(DD)andtypically
developing(TD)groups.Thefiguredisplaysmeansandstandarderrorsforthethreelearningstages.
3.2. Accuracyanalyses
The two groups performed at similar accuracy levels across the task as a whole (DD mean=87.9%, SD=15.2%; TD
mean=92.4%,SD=2.73%;maineffectofgroup:F(1,27)=1.53,p=.227,h 2=.053;group(cid:4)learningstageinteraction:F(2,
p
54)<1).Accuracylevelsdidnotchangesignificantlywithpractice(maineffectoflearningstage:F(1,27)=2.51,p=.091,
h 2=.085).
p
Fig.2Aplotsthemeanaccuracyforpatternandrandomtrialsasafunctionoflearningstage,andFig.2Bplotsthemean
accuracydifferencebetweenpatternandrandomtrialsasafunctionoflearningstage.ConsistentwithFig.2A,evidencefor
sequence learning, across all children, was found in that accuracy was higher for pattern (mean=91.7%, SD=10.1%)
comparedtorandom(mean=89.6%,SD=9.77%)trials(maineffectoftrialtype:F(1,27)=32.7,p<.0001,h 2=.548).The
p
accuracydifferencebetweenpatternandrandomtrialsincreasedasafunctionoflearningstage(trialtype(cid:4)learningstage
interaction:F(1,27)=7.04,p=.002,h 2=.207),suggestingthatparticipantsbecameincreasinglysensitivetothesequence
p
regularity with practice. Most importantly, however, there was a group difference in sequence learning (group(cid:4)trial
type(cid:4)learningstageinteraction:F(1,27)=3.84,p=.028,h 2=.124).
p
The triple interaction was followed-up with between-groups one-way ANOVAs with the pattern-random accuracy
differenceforeachlearningstageasthedependentvariable.TheseanalysesconfirmedthepatternsuggestedbyFig.2B.
TherewasatrendtowardslesssequencelearningintheDDgroupinthefirstlearningstage(maineffectofgroup:F(1,
27)=3.82,p=.061,h 2=.124).Atstage2,theamountofsequencelearningwashighlysimilarinthetwogroups(maineffect
p
ofgroup:F(1,27)<1).Asignificantgroupdifferenceemergedatstage3,withpoorerlearningintheDDgroupcomparedto
theTDgroup(maineffectofgroup:F(1,27)=4.38,p=.046,h 2=.139).
p
In sum, the accuracy analyses revealed no significant group differences in the first practice session (encompassing
learningstages1and2),althoughtherewasatrend(p=.061)towardslesslearningintheDDgroupcomparedtotheTD
groupinstage1.Lateinlearning,bycontrast,asignificantgroupdifferenceemerged.Bylearningstage3,afterextended
practiceandincludinganovernightinterval,theDDgroupshowedsignificantlylesssequencelearningcomparedtotheTD
group.
3.2.1. By-blockovernightconsolidationvsextendedpracticeeffects
Giventhattheobservedaccuracy-basedgroupdifferenceinlatelearning(stage3)couldbeduetodifferencesinovernight
consolidationand/orintheeffectsofextendedpractice,additionalanalyseswereperformedtodissociatetheseeffects.
Apotentialgroupdifferenceinovernightconsolidationwasexaminedbycomparingaccuracyonthefinalblock(i.e.10
repetitionsofthesequentialpattern)onthefirstday(block20)andthefirstblockonthesecondday(block21)witha2
(group:DDvsTD)(cid:4)2(trialtype:patternvsrandom)(cid:4)2(block:20vs21)ANOVA.ConsistentwithFig.3,whichsuggests
similarperformanceinthetwogroupsinbothblock20and21,thisanalysisrevealednogroupdifferencesinovernight
consolidationofthelearnedsequence(group(cid:4)trialtype(cid:4)blockinteraction:F(1,27)<1).Therewasevidenceforsequence
learningintheformofasignificantmaineffectoftrialtype(F(1,27)=7.40,p=.011,h 2=.215),reflectingoverallhigher
p
accuracyforpatterncomparedtorandomtrials,withasimilareffectinbothgroups(group(cid:4)trialtypeinteraction:F(1,
27)<1).Therewerenoindicationsofoff-linesequencelearningeffects,astheaccuracydifferencebetweenpatternand
randomtrialsdidnotincreaseovernight(trialtype(cid:4)blockinteraction:F(1,27)<1).However,overallaccuracyincreased
significantlyovernight(maineffectofblock:F(1,27)=20.4,p=.0001,h 2=.430),possiblyreflectingfatigueeffectsinthe
p
finalblockofsession1.Theovernightincreaseinoverallaccuracywasalsoverysimilarinthetwogroups(group(cid:4)block
interaction:F(1,27)<1).
[(Fig._3)TD$FIG]
3930 M.Hedeniusetal./ResearchinDevelopmentalDisabilities34(2013)3924–3935
Fig.3.Meanaccuracyforpatternandrandomtrialsasafunctionofpracticeblocksforthedevelopmentaldyslexia(DD)andtypicallydeveloping(TD)groups.
Thefiguredisplaysmeansandstandarderrorsforthefinalblockonday1(block20)andthe5blocksonday2(blocks21–25).Theverticallinebetweenblocks20
and21representsa24hintervalduringwhichnofurtherpracticewasgiven.Groupdifferenceswereexaminedforovernightconsolidation(blocks20–21)and
fortheeffectsofextendedpracticeonday2(blocks22–25).SeeSection3.2.1,foradetaileddescriptionoftheanalysesandresults.
Thelackofanysignificantgroupdifferencesintheovernightconsolidationofthelearnedsequencesuggeststhatthe
observedgroupdifferenceinlearningstage3maybeduetoadifferentialeffectoffurtherpracticeonday2inthetwogroups.
This prediction was supported by a group(cid:4)trial type(cid:4)block ANOVA for the remaining four blocks (blocks 22–25) in
learningstage3,againwithaccuracyasthedependentvariable(Fig.3).Inadditiontoasignificantmaineffectoftrialtype
(F(1, 27)=24.7, p<.0001, h 2=.478), this ANOVA produced a group(cid:4)trial type interaction (F(1, 27)=7.19, p=.012,
p
h 2=.210),reflectingasmallerpattern-randomaccuracydifference(i.e.sequencelearningeffect)intheDDgroupcompared
p
to theTD group(DD mean=0.0139, SD=0.0337, TDmean=0.0464, SD=0.0310). The lack of a group(cid:4)trial type(cid:4)block
interaction(F(3,81)<1)suggeststhattheTDadvantageremainedstableacrossthefinalfourblocks.Overallaccuracyand
reaction times did not differ between groups in either the consolidation analysis or the extended practice analysis (all
ps>.208).
3.3. Correlationbetweenimplicitsequencelearningandindividualreadingability
Next,therelationshipbetweenimplicitsequencelearningandindividualreadingabilitywasexaminedwithacorrelation
analysis. A composite measure of reading ability was calculated from the standardized raw scores of the phonological
decodingandorthographicreadingtests(Table1).Thiscompositereadingscorewasplottedagainsttheaccuracydifference
[(Fig._4)TD$FIG]betweenpatternandrandomtrialsatthefinallearningstage(stage3).AscanbeseeninFig.4,thisanalysisrevealeda
Fig.4.Scatterplotshowingthecorrelationbetweenimplicitsequencelearning(pattern-randomaccuracydifference)atstage3andthecombinedword
readingscore.
M.Hedeniusetal./ResearchinDevelopmentalDisabilities34(2013)3924–3935 3931
significantpositivecorrelationbetweenimplicitsequencelearningandindividualreadingability,independentofdiagnostic
category(r=.470,p=.010,r2=.221).
BecauseinspectionofFig.4suggeststhatascoreinthelow-leftcornerofthescatterplotcouldpotentiallybedrivingthis
correlation(thepattern-randomaccuracydifferencescoreforthisindividualwas(cid:5)3.0SDfromthemeanofallchildren,and
(cid:5)2.4SDfromthemeanoftheDDgroup),werepeatedthecorrelationanalysiswiththischildexcluded.Thecorrelationwas
stillsignificant(r=.426,p=.024,r2=.182),thussuggestingarobustassociationbetweensequencelearningandreadingskill
asacontinuousvariable.
3.4. Controlanalyses
In order to exclude the possibility that the group difference in sequence learning was an effect of lower general
intellectual ability in the DD compared to the TD group (as indicated by the near-significant group difference in PIQ),
additionalanalyseswereperformedinwhichtheeffectofPIQwasstatisticallycontrolledfor.Specifically,PIQwasaddedasa
covariateina2(group)(cid:4)3(learningstage)ANCOVAperformedontheaccuracydifferencebetweenpatternandrandom
trials(i.e.asdepictedinFig.2B).Thisanalysisproducedanidenticalsignificancepatterncomparedtotheanalyseswithout
covariatesreportedabove(stage3maineffectofgroup:p<.05).Thus,thegroupdifferenceinsequencelearningwasnot
explainedbylowergeneralintellectualabilityintheDDgroup.Importantly,theresultsheldalsowhenexcludingthechild
identifiedasanoutlierinthecorrelationanalysisabove(stage3maineffectofgroup:p<.05).
4. Discussion
The present study examined previously untested aspects of procedural memory, specifically of implicit sequence
learning, in children with DD compared to typically developing (TD) control children. Our study went beyond earlier
researchinexamininglearningstagesbeyonda singlepracticesession,andby allowingfora continuousassessmentof
sequencespecificlearningseparatefromgeneralskilllearning.
ChildrenwithDDshowedaselectivedeficitinimplicitsequencelearningwhilegeneralskilllearningwasnormal.In
addition, although there was a trend towards poorer learning in the DD group in the first learning stage, the sequence
learning impairment became significant only after extended practice, including an overnight interval. Further analyses
suggestedthattheimpairmentmaynotberelatedtoovernightconsolidation,butrathertotheeffectoffurtherpracticeon
day2.
Controlanalysesrevealedthattheobservedgroupdifferenceinsequencelearningwasnotexplainedbylowergeneral
intellectual ability in the DD group. In addition, the fact that overall reaction times and general skill learning were
comparableinthetwogroupssuggeststhattheresultswerenotduetogroupdifferencesinfatigue,generalprocessing
speed,orattention.
SequencelearningintheASRTtaskistypicallyseenasincreasinglyfasterresponsestopatterncomparedtorandomtrials
(RT based learning measure) and/or an increasing number of errors on random trials while accuracy for pattern trials
typicallyremainsmorestable(accuracybasedlearningmeasure).Thepatternofanincreasingfrequencyoferrorsonthe
unpredictabletrialswithpracticeistypicalwhenprobabilisticsequencesareused.Participantsoftenreportthattheirfingers
seem to‘‘takeover’’,leadingthemto makemore‘‘oops’’errors. Unbeknownsttothem,however,theseerrorsreflect an
increasingsensitivitytothesequentialregularity.Inthepresentstudy,significanteffectsofreadingskillasacategoricaland
continuousvariablewerefoundfortheaccuracybasedlearningmeasureonly,andnotfortheRTbasedmeasure.Thisfinding
isinlinewithtwopreviousstudiesexaminingsequencelearninginadultswithDDusingeithertheASRTtask(Howardetal.,
2006)ortheTripletLearningTask,whichcontainsasimilarprobabilisticsequencestructure(Bennett,Romano,Howard,&
Howard,2008).Inbothstudies,theaccuracybasedmeasureofsequencelearningwasshowntobemorestronglyrelatedto
readingabilitycomparedtotheRTbasedmeasure.Interestingly,bothofthesestudiesonadultswithDDreportcorrelations
betweenindividualreadingskillandtheaccuracybasedlearningscoreattheendofthetask,thatarestrikinglysimilartothe
correlationobservedinthepresentstudyperformedonchildren.
Overall,thefindingssupportthenotionofanassociationbetweenproceduralmemorydeficitsandDD.Theresultsarein
linewithpreviousstudiesreportingimplicitsequencelearningimpairmentsinchildren(Jimenez-Fernandezetal.,2011;
Vicarietal.,2003,2005)andadults(Du&Kelly,2013;Howardetal.,2006;Menghinietal.,2006;Stoodleyetal.,2006)with
DD.Theyextendpreviousfindingsbyshowing,forthefirsttime,thatthesequencelearningimpairmentinchildrenwithDD
may be most pronounced in learning stages beyond a single practice session and the early, fast acquisition phase.
Interestingly,asimilarpatternofresultshaspreviouslybeenreportedforchildrenwithspecificlanguageimpairment,a
conditionwhichoftenco-occurswithDD(Hedeniusetal.,2011).
ThemajorityofpreviousstudiesofimplicitsequencelearninginDDhavefoundagroupdifferenceearlyinlearning,i.e.
withinasinglepracticesession.Ourstudydidnotreplicatethesefindingsinthatthegroupdifferenceinearlylearningwas
only marginallysignificant inour stage 1 (p=.061). However,it is possiblethat this reflects a lack of powerdue to the
relativelysmallnumberofparticipants.Nevertheless,thelackofsignificanteffectsearlyinlearningisinterestinginlightof
the inconsistent results of previous studies. Indeed, all of the three studies that have reported intact implicit sequence
learninginchildrenwithDD(Deroostetal.,2010;Menghinietal.,2010;Waberetal.,2003)havefocusedonarelatively
narrowpracticeinterval(rangingfrom24to104repetitionsofthesequentialpatterncomparedwiththe250repetitionsin
3932 M.Hedeniusetal./ResearchinDevelopmentalDisabilities34(2013)3924–3935
thepresentstudy)givenwithinasinglepracticesession.Inlightofthefactthatasignificantgroupdifferenceinthepresent
studyemergedonlyinday2,aftermorethan200repetitionsofthesequence,itispossiblethatgroupdifferencesmayhave
beenrevealedifthestudiesabovehadincludedawiderpracticeinterval.
Impairmentonanon-languagetask,suchastheoneusedinthepresentstudy,challengesanyaccountthatpositsadeficit
specifictophonologicalprocessingastheunderlyingcauseofDD(Snowling,2000;Stanovich,1988;Vellutinoetal.,2004).
Per definition, any such account would predict intact non-language learning in DD, or at least, would not predict any
correlationbetweennon-languagesequencelearningandindividualreadingskill,aswasfoundhere.Thepresentfindings
are thus in line with data suggesting that the neurocognitive deficits in children with DD may go beyond a selective
phonologicalimpairment(seeSection1).
OurfindingsarecompatiblewiththeproceduralmemorydeficitviewofDD(Nicolsonetal.,2001,2010;Ullman,2004).
The findings may also be compatible with the magnocellular theory of DD (Stein, 2001), since this theory includes the
predictionthatcerebellarfunctions,andthusaspectsofproceduralmemory,areimpairedinthedisorder.Notably,thefact
thatasignificantDDimpairmentemergedrelativelylateinlearningmaypointtoastriatalratherthan,orinadditionto,a
cerebellar dysfunction (Doyon et al., 2009). As noted in the introduction, the relative contribution of different brain
structures/circuitstosequencelearninghasbeenshowntochangeaslearningprogressesthroughthedifferentstages.Brain
imagingstudiessuggestthatwhilethestriatum,thecerebellum,andthemedialtemporallobeshowsimultaneouslearning-
dependentactivityinthefastacquisitionstage(withthemedialtemporallobeoftenbeingrecruitedveryearlyinlearning),
activationinthecerebellumtypicallydecreaseswithpracticeandisusuallynolongerdetectablewhenthesequenceiswell
learned(Doyonetal.,2002,2009;Jueptner,Frith,Brooks,Frackowiak,&Passingham,1997;Rieckmann&Backman,2009;
Rieckmannetal.,2010;Simonetal.,2012).Thestriatum,bycontrast,remainsactivatedthroughtheasymptoticlearningas
wellastheconsolidationandlatelearningstages(Doyon,Owen,Petrides,Sziklas,&Evans,1996;Lehericyetal.,2005).This
patternofbrainactivationhasbeentakenasevidenceforanincreasinglycriticalroleofthestriatuminthelearningstages
beyondthefastacquisitionstage(Bennett,Madden,Vaidya,Howard,&Howard,2011;Doyonetal.,2009).However,any
claimsabouttheneuralbasisoftheobservedsequencelearningdeficitinthepresentstudyarespeculative.
Even though the majority of studies to date support the notion of an association between poor reading and implicit
sequence learning deficits, the nature of this association remains to be elucidated. The association revealed here, and
elsewhere,mayreflectoneofseveralpossiblerelationships.Onepossibilityisthatageneralproceduralmemorydeficit
underlies the core phonological problems in DD as well as a range of previously reported cognitive and motor deficits
(Nicolsonetal.,2010),includingthatofimplicitsequencelearningobservedinthepresentstudy.Thisistheproposalofthe
proceduralmemorydeficitview(Nicolsonetal.,2001,2010;Ullman,2004).Onthisview,aproceduralmemorydeficitmay
affectreadingbothdirectly,throughskillautomatizationproblems,andindirectly,viaphonologicalprocessingproblems
(Nicolson&Fawcett,2011).
Anotherpossibilityisthatreadingandproceduralmemorydeficitstendtoco-occurinDDbutarecausallyunrelated.That
is,forreasonsyetunknown,aproceduralmemorydeficitmaybemorecommonindyslexicchildrencomparedtotheir
typicallydevelopingpeers,butthisimpairmentdoesnotcausethereadingproblems,oraffecttheminanysubstantialway.
Such a view is supported by evidence suggesting that sensori-motor impairments, such as those predicted by both the
proceduraldeficitandmagnocellularviews,mayneitherbenecessarynorsufficienttocausethereadingproblemsobserved
inDD(Ramus,2003;Whiteetal.,2006).However,whilethisviewiscompatiblewiththeobservedgroupdifferencein
sequencelearning,itissomewhatdifficulttoreconcilewiththelinearrelationshipbetweenreadingandimplicitsequence
learning,independentofdiagnosticcategory,observedhere,andinotherstudies(Bennettetal.,2008;Howardetal.,2006).
Athirdpossibilityisthatthecorephonologicaldeficitandtheproceduralmemorydeficitarecausallyunrelatedbutthat
theco-occurrenceofthesedeficitsmayleadtoreadingproblemsthataresevereenoughtodrawpedagogicalandclinical
attention.Onthisview,procedurallearningproblemsmaynotcausethephonologicalimpairmentinDDbutmaycontribute
independentvariancetoindividualreadingabilitybyaffecting,forexample,theabilitytoextractorthographicsequential
regularities and the automatization of reading skills. Such a ‘‘double deficit’’ view has previously been proposed for
phonologicalprocessingproblemsandrapidnamingability(Wolf&Bowers,1999).Thisviewwouldbeinlinewithstudiesof
childrenatriskfordyslexia,inwhichithasbeenshownthatphonologicalproblemsalonearenotsufficienttocausesevere
readingproblems(Gallagher,Frith,&Snowling,2000;Snowlingetal.,2003;Wimmer&Schurz,2010).
Thedesignofthepresentstudydoesnotallowustospecifythenatureoftherelationshipbetweenproceduralmemory
andreadingskill,butonlytodemonstratethatsucharelationshipexists.Futurestudies,withalongitudinaldesignstarting
atbirth,andincludingawiderrangeoflinguistic,cognitiveandmotortasks,mayhelpshedlightonthisissue.
Thepresentstudyhasvariousmethodologicallimitationsthatmaybeaddressedbyfutureresearch.Onesuchlimitation
isthatsamplesizeswererelativelysmall.Asnotedabove,thismayhaveaffectedthepowertodetectagroupdifferencein
earlylearning.Thus,thepresentfindingsneedtobereplicatedwithlargersamplesinordertoensuretheirgeneralizability.
Inaddition,futurestudiesofproceduralmemoryconsolidationandextendedpracticeeffectsinchildrenwithDDwould
benefitfromincludingamoredirecttestofconsolidation(suchastheamountofinterferencebyaneworrandomsequence;
Robertsonetal.,2004),aswellasmorelearningtrialsfortheexaminationoffurtherpracticeonday2.Suchparadigmsmay
helpshedfurtherlightontheDDdeficitinlatelearningdemonstratedhere.
Futurestudieswouldalsobenefitfromneuroimagingdatatosupporttheassumptionthatchildrenrelyonprocedural
memory for sequence learning in the ASRT task. Although previous fMRI-studies have indeed revealed activation of
proceduralmemorybrainstructuresassociatedwithASRTlearninginadults(Bo,Peltier,Noll,&Seidler,2011),thelackof
M.Hedeniusetal./ResearchinDevelopmentalDisabilities34(2013)3924–3935 3933
datafromstudiesofchildrenperformingthistaskprecludesanystrongconclusionsregardingbraincorrelatesoflearningin
thisstudy.
Keepingtheselimitationsinmind,thepresentfindingslendsupporttothenotionthatDDisassociatedwithaprocedural
memorydysfunction.Further,theresultsemphasizetheimportanceofstudyingnotonlyinitiallearning,butalsomemory
consolidation and longer-term learning in children with DD in order to obtain a deeper understanding of learning and
memoryfunctionsinaffectedchildren.Suchknowledgewouldpotentiallybeofgreatimportancetothedevelopmentof
effectivepedagogicalandtherapeuticremediationstrategies.
Acknowledgments
This workwas supportedby grants from theSven Jerring Foundationand the SunnerdahlDisability Foundation.We
gratefullyacknowledgethehelpofthemanyschoolsandfamilieswhoparticipatedinthisstudy.
References
Albouy,G.,Sterpenich,V.,Balteau,E.,Vandewalle,G.,Desseilles,M.,Dang-Vu,T.,etal.(2008).Boththehippocampusandstriatumareinvolvedinconsolidationof
motorsequencememory.Neuron,58,261–272.
Aldridge,J.W.,&Berridge,K.C.(1998).Codingofserialorderbyneostriatalneurons:Anaturalactionapproachtomovementsequence.JournalofNeuroscience,18,
2777–2787.
Bennett,I.J.,Madden,D.J.,Vaidya,C.J.,Howard,J.H.J.,&Howard,D.V.(2011).Whitematterintegritycorrelatesofimplicitsequencelearninginhealthyaging.
NeurobiologyofAging,32,2317.e1–2317.e12.
Bennett,I.J.,Romano,J.C.,Howard,J.H.J.,&Howard,D.V.(2008).Twoformsofimplicitlearninginyoungadultswithdyslexia.AnnalsoftheNewYorkAcademyof
Sciences,1145,184–198.
Bishop,D.V.(1982).T.R.O.GTestforreceptionofgrammar.Manchester:ChapelPress.
Bo,J.,Peltier,S.J.,Noll,D.C.,&Seidler,R.D.(2011).Symbolicrepresentationsinmotorsequencelearning.Neuroimage,54,417–426.
Brosnan,M.,Demetre,J.,Hamill,S.,Robson,K.,Shepherd,H.,&Cody,G.(2002).Executivefunctioninginadultsandchildrenwithdevelopmentaldyslexia.
Neuropsychologia,40,2144–2155.
Conway,C.M.,Bauernschmidt,A.,Huang,S.S.,&Pisoni,D.B.(2010).Implicitstatisticallearninginlanguageprocessing:Wordpredictabilityisthekey.Cognition,
114,356–371.
Conway,C.M.,&Pisoni,D.B.(2008).Neurocognitivebasisofimplicitlearningofsequentialstructureanditsrelationtolanguageprocessing.AnnalsoftheNew
YorkAcademyofSciences,1145,113–131.
Debas,K.,Carrier,J.,Orban,P.,Barakat,M.,Lungu,O., Vandewalle,G.,etal.(2010).Brainplasticityrelatedtotheconsolidationofmotorsequencelearningand
motoradaptation.ProceedingsoftheNationalAcademyofSciencesoftheUnitedStatesofAmerica,107,17839–17844.
Deroost,N.,Zeischka,P.,Coomans,D.,Bouazza,S.,Depessemier,P.,&Soetens,E.(2010).Intactfirst-andsecond-orderimplicitsequencelearninginsecondary-
school-agedchildrenwithdevelopmentaldyslexia.JournalofClinicalandExperimentalNeuropsychology,32,561–572.
Dominey,P.F.,Hoen,M.,Blanc,J.M.,&Lelekov-Boissard,T.(2003).Neurologicalbasisoflanguageandsequentialcognition:Evidencefromsimulation,aphasia,
andERPstudies.BrainandLanguage,86,207–225.
Doyon,J.,Bellec,P.,Amsel,R.,Penhune,V.B.,Monchi,O., Carrier,J.,etal.(2009).Contributionsofthebasalgangliaandfunctionallyrelatedbrainstructuresto
motorlearning.BehaviouralBrainResearch,199,61–75.
Doyon,J.,Owen,A.M.,Petrides,M.,Sziklas,V.,&Evans,A.C.(1996).Functionalanatomyofvisuomotorskilllearninginhumansubjectsexaminedwithpositron
emissiontomography.EuropeanJournalofNeuroscience,8,637–648.
Doyon,J.,Song,A.,Karni,A.,Lalonde,F.,Adams,M.,&Ungerleider,L.(2002).Experience-dependentchangesincerebellarcontributionstomotorsequence
learning.ProceedingsoftheNationalAcademyofScienceoftheUnitedStatesofAmerica,99,1017–1022.
Du,W.,&Kelly,S.W.(2013).Implicitsequencelearningindyslexia:Awithin-sequencecomparisonoffirst-andhigher-orderinformation.AnnalsofDyslexia,63,
154–170.
Dunn,L.M.,&Dunn,L.M.(1997).Peabodypicturevocabularytest(3rded.).CirclePines:AmericanGuidanceService.
Ferrer,E.,Shaywitz,B.A.,Holahan,J.M.,Marchione,K.,&Shaywitz,S.E.(2010).UncouplingofreadingandIQovertime:Empiricalevidenceforadefinitionof
dyslexia.PsychologicalScience,21,93–101.
Gabrieli,J.D.(1998).Cognitiveneuroscienceofhumanmemory.AnnualReviewofPsychology,49,87–115.
Gallagher,A.,Frith,U.,&Snowling,M.J.(2000).Precursorsofliteracydelayamongchildrenatgeneticriskofdyslexia.JournalofChildPsychologyandPsychiatry,41,
203–213.
Gheysen,F.,VanOpstal,F.,Roggeman,C.,VanWaelvelde,H.,&Fias,W.(2011).Theneuralbasisofimplicitperceptualsequencelearning.FrontiersinHuman
Neuroscience,5,137.
Grafton,S.T.,Hazeltine,E.,&Ivry,R.(1995).Functionalmappingofsequencelearninginnormalhumans.JournalofCognitiveNeuroscience,7,497–510.
Hari,R.,&Renvall,H.(2001).Impairedprocessingofrapidstimulussequencesindyslexia.TrendsinCognitiveSciences,5,525–532.
Hauptmann,B.,&Karni,A.(2002).Fromprimedtolearn:Thesaturationofrepetitionprimingandtheinductionoflong-termmemory.CognitiveBrainResearch,13,
313–322.
Hauptmann,B.,Reinhart,E.,Brandt,S.A.,&Karni,A.(2005).Thepredictivevalueofthelevelingoffofwithin-sessionperformanceforproceduralmemory
consolidation.CognitiveBrainResearch,24,181–189.
Hedenius,M.,Persson,J.,Tremblay,A.,Adi-Japha,E.,Verı´ssimo,J., Dye,C.D.,etal.(2011).Grammarpredictsprocedurallearningandconsolidationdeficitsin
childrenwithspecificlanguageimpairment.ResearchinDevelopmentalDisabilities,32,2362–2375.
Henke,K.(2010).Amodelformemorysystemsbasedonprocessingmodesratherthanconsciousness.NatureReviewsNeuroscience,11,523–532.
Hensler,B.S.,Schatschneider,C.,Taylor,J.,&Wagner,R.K.(2010).Behavioralgeneticapproachtothestudyofdyslexia.JournalofDevelopmentalandBehavioral
Pediatrics,31,525–532.
Holmberg,E.,&Lunda¨lv,E.(2002).T.R.O.G.svenskmanual.Go¨teborg:SpecialpedagogiskainstitutetLa¨romedel.
Howard,D.V.,Howard,J.H.,Jr.,Japikse,K.C.,DiYani,C.,Thompson,A.,&Somberg,R.(2004).Implicitsequencelearning:Effectsoflevelofstructure,adultage,and
extendedpractice.PsychologyandAging,19,79–92.
Howard,D.V.,&Howard,J.H.,Jr.(1992).Adultagedifferencesintherateoflearningserialpatterns:Evidencefromdirectandindirecttests.PsychologyandAging,
7,232–241.
Howard,J.H.,Jr.,&Howard,D.V.(1997).Agedifferencesinimplicitlearningofhigher-orderdependenciesinserialpatterns.PsychologyandAging,12,634–656.
Howard,J.H.,Jr.,Howard,D.V.,Japikse,K.C.,&Eden,G.F.(2006).Dyslexicsareimpairedonimplicithigh-ordersequencelearning,butnotonimplicitspatial
contextlearning.Neuropsychologia,44,1131–1144.
Jimenez-Fernandez,G.,Vaquero,J.M.,Jimenez,L.,&Defior,S.(2011).Dyslexicchildrenshowdeficitsinimplicitsequencelearning,butnotinexplicitsequence
learningorcontextualcueing.AnnalsofDyslexia,61,85–110.