La lecture à portée de main
Description
Sujets
Informations
Publié par | universitat_des_saarlandes |
Publié le | 01 janvier 2005 |
Nombre de lectures | 15 |
Langue | English |
Poids de l'ouvrage | 9 Mo |
Extrait
The
Neural
Bases
of
yRegularit
Dissertation
Learning
zurErlangungdesGradeseinesDoktorsderPhilosophie
der
PhilosophischenFakultatenderUniversitatdesSaarlandes
vorgelegtvonDipl-Psych.ChristianDoller
orengeb
amin10.12.1973
urzburgW
an:Dek
Prof.Dr.RainerKrause,UniversitatdesSaarlandes
terstatter:hBeric
Prof.Dr.AxelMecklinger,UniversitatdesSaarlandes
PDZimmer,ertHubDr.
TDisputation:derag
14.02.2005
atersitUniv
des
Saarlandes
I
urF
anjaT
I
I
♥
Wemustsupposeaverydelicateadjustmentwherebythecirculationfollowstheneedsof
thecerebralactivity.Bloodverylikelymayrushtoeachregionofthecortexaccordingas
itismostactive,butofthisweknownothing.
WilliamJames,ThePrinciplesofPsychology(1890)
Cognitiveneurosciencecanmoveforwardwithgreatercondenceintheknowledgethat
changesinbloodowandoxygenlevelsdorepresentdenablealterationsinneuronalac-
tivity.
III
MarcusE.Raichle,Nature,412(2001)
Abstract
Theexistenceoftwoseparatelearningandmemorysystemshasbeenproposedinexperi-
mentalpsychologyandcognitiveneuroscience(Chapter1),specializedforcomplementary
functions,namely(1)fastlearninguniqueepisodes(rstsystem;Chapter2)and(2)grad-
uallearningregularitiesacrossmultipleepisodes(secondsystem;Chapter3).Thepresent
thesisaimedatspecifyingtheneuralandcognitivebasesofthesecondsystem.Within
theframeworkofthisthesis,theterm‘regularitylearning’wasintroducedtodenethe
learningprocessassociatedwiththesecondsystem,namelytheextractionofregularities,
i.e.overlapping,invariantfeaturesofmultipleepisodes.
Theneuralandcognitiveprocessesunderlyingregularitylearningwereinvestigatedin
threeexperiments.Twobrainregionswereofmaininterest,thehippocampusandthe
prefrontalcortex,PFC(Chapter4).Functionalmagneticresonanceimaging(fMRI)was
usedasthemainmethod(Chapter5).Inallexperiments,volunteershadtolearnobject-
positionconjunctionsinseveralexperimentaltrialsandblocks.Thedesignoftheexper-
imentsincludedtwoconditions,acontext-specic(CS)andaninvariantlearning(IL)
condition(Chapter6).IntheCScondition,objectsandpositionswerevariablymapped
acrosstrials.Incontrast,intheILconditionpositions(Experiment1Aand3B)orobjects
(Experiment1B,2,and3A)wereheldconstantwithinblocks,enablingsubjectstoextract
regularitiesacrosstrials,i.e.invariantpositionsorobjectsinobject-positionconjunctions,
.elyectivrespInExperiment1A(Chapter7)andExperiment1B(Chapter8),performanceincreased
acrosstrialswithinblocksoftheILcondition,butnotintheCScondition.Inbothex-
periments,hippocampalactivitydecreasedasafunctionoflearningintheILcondition.
Conversely,thehippocampuswasactivatedcontinuouslyintheCScondition.Incontrast
toaright-lateralizedhippocampalactivationdecreaseduringlearningspatialregularities
(Experiment1A),learningobjectregularities(Experiment1B)wasassociatedwithade-
creaseofbilateralhippocampalactivity.Inbothexperiments,leftlateralprefrontaland
rightstriatalbrainregionsshowedanincreaseofactivityasafunctionoflearninginthe
ILcondition.InExperiment2(Chapter9),thecognitiveprocessesunderlyinglearning
objectregularitieswereinvestigatedinmoredetail.Insixbehavioralstudies,theinuence
ofobjectdistinctiveness(Experiment2A-2B),memoryloadandlearningduration(Exper-
iment2C-2E),andtheeectofsimultaneousobject-positionbindings(Experiment2F)on
learninginvariantobjectsinobject-positionconjunctionswasexamined.Thebehavioral
IV
V
dataindicatethattheincreaseofobjectdistinctivenessmainlyaectsoverallmemoryper-
formance,whereasmanipulationofthelearningdurationandthememoryloadaectsthe
learningprocess.Moreover,adissociationoftwolearningmechanismscouldbeobserved:
aprocessoperatingwithinlearningblocksandaprocessoperatingacrossblocks.Finally,
Experiment3(Chapter10)aimedatspecifyingwhetherseparablebrainregionsmediate
thesetwolearningmechanisms.IncontrasttoExperiment1Aand1B,trialswereblocked
byconditiontominimizetheprobabilitythatsubjectsbasedtheirjudgmentonacom-
monstrategyforbothconditions.Onabehaviorallevel,subjectsshowedaperformance
increasewithinandacrosslearningblocksinanobjectILcondition(Experiment3A)
andinaspatialILcondition(Experiment3B),butnotintherespectiveCSconditions.
Onaneurallevel,within-blocklearningwasassociatedwithalearning-relateddecrease
ofhippocampalandalearning-relatedincreaseofprefrontal-striatalactivityinbothex-
periments,bythisreplicatingtheresultsofExperiment1Aand1B.Incontrast,distinct
prefrontal-striatalregionswereselectivelyinvolvedinacross-blocklearning.
Tosummarize,onacognitivelevelregularitylearningwithintheframeworkofthepresent
thesiscanbecharacterizedbythefollowingcriteria:(1)theextractionofregularities
acrossinputpattern,(2)thegradualnatureoftheextractionprocess,(3)therobust
maintenanceofextractedregularitiesovertime,and(4)theaggregatedrepresentational
formatoftheextractedinformation.Furthermore,thedistinctionbetweenawithin-block
andanacross-blocklearningeectmightindicatethattheformereectisassociatedwith
aninstance-basedlearningprocess,whereasthelattereectmightreectarule-based
learningprocess,i.e.thetransferofknowledgeofinvariantfeaturestonewinstances.On
aneurallevel,regularitylearningwasassociatedwithadecreaseofhippocampalandan
increaseofprefrontal-striatalactivity.Thereducedlearning-relatedhippocampalactiva-
tionpresumablyreectslowerbindingrequirementsintheILcondition,sincevariable
objectscanbeboundtoinvariantpositions(Experiment1Aand3)orviceversa(Experi-
ment1B,2,and3).Furthermore,theimagingresultsofExperiment1Aand1Bsuggesta
domain-specichemisphericspecializationofthehippocampusduringregularitylearning,
reectinghippocampalsensitivitytoperceptualstimulusattributesofinvariantepisodic
features(bottom-upmechanism).Thelearning-relatedprefrontalmodulationseemstore-
ecttherequirementtoextractandmaintainregularitiesacrosstrialsandtheadjustment
ofobject-positionconjunctionsonthebasisoftheextractedknowledge,possiblymediated
byrule-likeprefrontalrepresentations(top-downmechanism).Finally,thestriatummight
encodetheincreasedpredictabilityofinvariantfeaturesasafunctionoflearningandpos-
siblyprovidesaninternalreinforcementsignaltothePFC.
Inconclusion,thepresentresultsprovidenewinsightsintotheneuralbasisofregularity
learningandpointtoatransitionoftherelativerolesofdistinctneuralsystemsdur-
ingthetime-courseofregularitylearning,i.e.learningisaccompaniedbyashiftfroma
hippocampaltoaprefrontal-striatalbrainsystem.
tstenCon
erviewOv0
troInIduction
ductiontroIn1
1
4
5
10TheoryII2FastLearningUniqueEpisodes:TheHippocampusandRelationalBind-
11ing2.1Introduction....................................12
2.2RelationalBindingandtheHippocampus:TheEichenbaumModelandthe
O’ReillyModel..................................14
2.2.1TheEichenbaumModel.........................14
2.2.2TheO’ReillyModel...........................15
2.3StudiesfromDierentAreasofBehavioralandCognitiveNeuroscience...16
2.3.1SingleCellRecordingsandLesionStudiesinAnimals........17
2.3.2AmnesiaResearch............................20
2.3.3ElectrophysiologicalStudiesinHumans................23
2.3.4ImagingStudies.............................26
2.4RepresentationofRegularitiesacrossMultipleEpisodes...........31
2.4.1AssumptionsoftheEichenbaumModel................31
2.4.2AssumptionsoftheO’ReillyModel...................32
2.4.3ConcludingRemarks...........................33
3GradualLearningRegularitiesacrossMultipleEpisodes:BasicPrinci-
plesandNeuralMechanisms35
3.1Introduction....................................36
3.2ReinforcementLearning.............................38
3.2.1BasicPrinciples..............................38
3.2.2NeuralMechanisms............................39
VI
CONTENTS
IVI
3.2.3Summary.................................46
3.3CategoryLearning................................46
3.3.1BasicPrinciples..............................46
3.3.2NeuralMechanisms............................47
3.3.3Summary.................................51
3.4ArticialGrammarLearning..........................51
3.4.1BasicPrinciples..............................51
3.4.2Rule-basedvsSimilarity-basedLearning................54
3.4.3NeuralMechanisms............................55
3.4.4Summary.................................57
3.5SequenceLearning................................57
3.5.1BasicPrinciples..............................57
3.5.2NeuralMechanisms............................62
3.5.3Summary.................................65
4NeuroanatomicalBasis:TheMedialTemporalLobeandthePrefrontal
66Cortex4.1MedialTemporalLobe..............................67
4.1.1MainStructures.............................67
4.1.2MainPathways..............................68
4.2PrefrontalCortex.................................69
4.2.1MainStructures.............................69
4.2.2MainPathways..............................70
4.3MedialTemporal-prefrontalInterconnections.................72
dsMethoIII
74
75NeuroimagingunctionalF55.1PrinciplesofMagneticResonanceImaging...................76
5.1.1NuclearSpinSystemsintheMagnet..................76
5.1.2ExcitationoftheSpinSystem......................77
5.1.3MagneticResonanceImaging......................80
5.1.4ImagingParametersandSequences...................81
5.2TheBOLDSignal................................83
5.2.1PhysicalBasis..............................83
5.2.2PhysiologicalBasis..