Neural
systems
supporting
navigation
Hugo
J
Spiers
1and
Caswell
Barry
2Muchisknownabouthowneuralsystemsdeterminecurrent
spatialpositionandorientationintheenvironment.Bycontrast
littleisunderstoodabouthowthebrainrepresentsfuturegoal
locationsorcomputesthedistanceanddirectiontosuchgoals.
Recentelectrophysiology,computationalmodellingand
neuroimagingresearchhaveshednewlightonhowthespatial
relationshiptoagoalmaybedeterminedandrepresentedduring
navigation.Thisresearchsuggeststhatthehippocampusmay
codethepathtothegoalwhiletheentorhinalcortexrepresents
thevectortothegoal.Italsorevealsthattheengagementofthe
hippocampusandentorhinalcortexvariesacrossthedifferent
operationalstagesofnavigation,suchasduringtravel,route
planning,anddecision-makingatwaypoints.
Addresses
1InstituteofBehaviouralNeuroscience,DepartmentofExperimental
Psychology,DivisionofPsychologyandLanguageSciences,UCL, 26BedfordWay,LondonWC1H0AP,UK
2
DepartmentofCellandDevelopmentalBiology,UCL,GowerStreet, LondonWC1E6BT,UK
Correspondingauthor:Spiers,HugoJ([email protected])
CurrentOpinioninBehavioralSciences2015,1:47–55
ThisreviewcomesfromathemedissueonCognitiveneuroscience EditedbyCindyLustigandHowardEichenbaum
ForacompleteoverviewseetheIssueandtheEditorial Availableonline6thSeptember2014
http://dx.doi.org/10.1016/j.cobeha.2014.08.005
2352-1546/#2014TheAuthors.PublishedbyElsevierLtd.Thisisan openaccessarticleundertheCCBYlicense( http://creativecom-mons.org/licenses/by/3.0/).
Introduction
Theabilitytonavigateisafundamentalbehaviourshared
bymostmotileanimalsonourplanet.Inordertonavigate
ananimalmustdeterminethedirectiontotravelin,how
farto traveland subsequentlykeep trackof itsprogress
through theenvironment.The challengesof navigating
varydependingontheenvironment.Forexample,
navi-gating an open featureless terrain presents different
challenges to traversing an urban street network.
Sim-ilarly,recallingwherealocationisandhowtogetthereis
likelytobemorechallenginginanovelenvironmentthan
a well-known one. When navigation requires travelling
along familiar habitual routes evidence indicates that
stimulus–responseassociationsstoredin thedorsal
stria-tumallow ananimalto determineinwhich directionto
proceedandwhentheyhavetravelledfarenoughtoarrive
at the goal [1–3]. However, when navigation relies on
determiningself-locationintheenvironmentand
comput-ingthe spatialrelationshipto thegoal, thehippocampus
and connected structures of the medial temporal lobe
(MTL), such as the entorhinal cortex, are needed for
navigation[4–8].MTL andstriatum alsooperateas part
ofawiderbrainnetworkservingnavigation.Insummary,it
isthoughttheparahippocampalcortexsupportsthe
recog-nitionofspecificviewsandtheretrosplenialcortexconverts
betweenallocentric(environment-bound)representations
in hippocampal–entorhinal regions to egocentric
repres-entationsinposteriorparietalcortex[9,10,11].Inaddition,
the prefrontal cortex is thought to aid route planning,
decision-makingandswitchingbetweennavigation
strat-egies[12,13]andthecerebellumisrequiredwhen
naviga-tioninvolvesmonitoringself-motion[14].Herewefocuson
theroleofthehippocampusandentorhinalcortexbecause
ofrecentdiscoveriesfromfunctionalmagnetic resonance
imaging(fMRI)andsingleunitrecordingstudiesandthe
developmentofnewcomputationalmodels.
Electrophysiologicalinvestigationshaverevealedseveral
distinct neural representations of self-location (see
Figure 1 and forreview [15]).Briefly, place cellsfound
inhippocampalregionsCA3andCA1signaltheanimal’s
presence in particular regions of space; the cells’place
fields [16] (Figure 1a). Place fields are broadly stable
betweenvisitstofamiliarlocationsbutremapwhenever
a novelenvironment is encountered, quickly forming a
newanddistinctrepresentation[17,18].Gridcells,
ident-ified in entorhinal cortex,and subsequently in the
pre-subiculum and para-subiculum, also signal self-location
but dosowithmultiplereceptive fieldsdistributedin a
strikinghexagonalarray[19,20](Figure1b).Head
direc-tioncells,foundthroughoutthelimbicsystem,providea
complementary representation, signalling facing
direc-tion; with eachcell responding only when theanimal’s
head is within a narrow range of orientations in the
horizontalplane(e.g.[21],Figure1c). Othersimilarcell
types are also known, for example border cells which
signal proximity to environmental boundaries [22] and
conjunctivegridcellswhichrespondtobothpositionand
facingdirection[23].Itislikelythatthesespatial
repres-entationsareacommonfeatureofthemammalianbrain,
attheveryleastgridcellsandplacecellshavebeenfound
in animalsas diverseas bats,humans, androdents[15].
Goal-related
coding
in
spatial
cells
Do these representations of self-location play a role in
guidingnavigation?Theactivityofspatialneuronsco-varies
errorintheheaddirectionsystempredictsthebearingrats
takewhenattempting toreachagoal[24,25]andsimilar
resultsareknownforplacecells[26].Whilethissuggests
thatthesecellsformthebasisofnavigationalcomputations
itisnotclearwhatformthosecomputationstakeandwhere
theyaremade.Inparticular,howspatialnetworksencode
goallocationandutilisethisinformationtodeterminean
appropriateroutearestilltobedetermined.However,the
lastdecadehasseensomeprogresswiththeformerofthese
problems. For example, it isnow known thatplace cell
populationsencodeinformationinadditiontothe
repres-entationofself-location,suchaspresenceofrewardatagoal
locations[27],ortherecentandfutureturnstobemadeina
route[28,29]. There have been conflicting reports as to
Figure1 orientation spacing offset (a) (b) (c) 3.2Hz neural data position tracking 26.8Hz 14.0Hz 42º 0º 180º 166º 0º 180º 5.6Hz
Current Opinion in Behavioral Sciences
Single-unitrecordingsofneuronsencodingaspectsofself-location.(a)CA1placecellrecordingmadefromarat.Left-handfigureshowsatypical experimentalsetup,theanimalforagesina1m2openfieldenvironment,inconcertitspositionistrackedbyanoverheadcameraandactionpotentials arerecordedfromimplantedelectrodes.Centre,rawdata:theblacklineindicatestheanimal’scumulativepathover20min;superimposedgreendots indicatingthelocationatwhichthiscellfiredactionpotentials.Right,thesamedataprocessedtoshowfiringrate(numberofspikesdividedbydwell time)perspatialbin.‘Hot’coloursindicatehighfiringratesand‘cold’colourslowfiringrates,whitebinsareunvisited,peakfiringrateisshownabove themap.(b)Toprow,rawdataandcorrespondingratemapforasinglemedialentorhinalcortex(mEC)gridcellshowingthemultiplefiringfields arrangedinahexagonallattice.Bottomrow,threeco-recordedmECgridcells,thecentreofeachfieldisindicatedbyacross,colourscorresponding todifferentcells.Thefiringpatternofeachcelliseffectivelyatranslationoftheotherco-recordedcellsasshownbytherelativepositionofthecrosses (right).(c)TwoheaddirectioncellsrecordedfrommECdeeplayers(V/VI).Firingrateisdisplayedasafunctionofheaddirection,thecellonthelefthas apeakfiringrateof26.8Hzachievedwhentheanimalwasfacingatanorientationof428(measuredanti-clockwisefromthehorizontalaxisofthe environment).
whether rodent hippocampal place cells preferentially
representgoallocations[12].Navigationinenvironments
composedof tracks(suchas T-mazesorplus-mazes)has
tendednotto findgoal-locationrelatedfiring [30,31].By
contrast,in open-fieldenvironments,whichmake greater
demands on self-localisation for navigation, elevated
place cell activity proximate to goals has been reported
[32,33–35].Similarly,theactivityofhippocampalcellsin
pre-surgicalepilepticpatientsnavigatinginavirtualtown
hasbeenshowntobemodulatedbythecurrentgoal[36].A
recent important study in which rats learned new goal
locations eachdayinanopenarena,foundthatCA1,but
not CA3,placecells, showed shifts in firing towardsthe
newlylearnedgoallocations [32].Cellsintheprelimbic
frontal cortex have also been reported to show activity
clusteredaroundgoallocationsinanopenarena.However,
no such clustering of activity near goal locations was
observed when ratscould relyona visualmarker ofthe
goal,ratherthantheirmemory,tolocatethegoal[35,37].
Computational
models
of
navigational
guidance
systems
Numerous computational models have sought to
under-standhownavigationcanbeconductedonthebasisofthe
knownorpredictedneuralrepresentations.Beforethe
dis-coveryofgridcellsthisworkwasprimarilyfocusedonplace
cells(e.g.[38–41]).However,becauseplacecellsexhibita
sparsespatialcodeofirregularfieldsitisnotobviousthat
theyencodethestructureoflargescalespace;theydonot
provideaspatialmetric[42].Inotherwords,basedonthe
population activity of place cells at two positions in the
environmentitisdoesnotappearthattherelativeproximity
ofthosepositionscanbeeasilyinferred.
Modelsaddressedthisissueinseveralways;one
possib-ility beingthat therelative proximity of place fields is
learntduringaperiodofexploration.Forexample,
Heb-bian-likeorspiketimedependentplasticitywilltendto
strengthen connectionsbetweenplace cells with
neigh-bouringfieldsbecauseadjacentlocationsarevisitedmore
frequently andinclosertemporalsequencethandistant
locations [40,41].In this way the connectivity between
place cells, normally identified with theCA3 recurrent
connections,isupdatedtoreflecttherelativepositionof
their fields in space and can be used to test or infer
potentialroutes[41].Aweaknessofthisapproachthough
isthattheanimalmustthoroughlyexploreanunfamiliar
environment before it can navigate effectively;
specifi-cally the network cannot identify routes that traverse
unvisited sections of space. Thus, the system cannot
exploitpotentialshortcutswhenchangestothe
environ-ment occur.Conversely, itdoes meanthatthe network
learnsabouttherelativeaccessibilityofpointsinknown
space,allowingtheshortestroutetobeselectedand
dead-endsavoided.Mulleretal.’s[41]modeloftheCA3place
cell network as a resistive grid took advantage of this
effecttodeterminetheshortestviableroutetoagoal.An
alternative proposalis thatnavigationcouldbeaffected
bymovingtomaximisethesimilaritybetweentheplace
cellrepresentationofthegoalandcurrentlocation.
How-ever,suchanapproachisonlysuccessfulwhentravelling
betweenpointsseparatedbylessthanthediameterofthe
largest place field. Beyond this distance the overlap
betweenrepresentationswillbeflataffordingnogradient
to follow.Althoughthesizeofthelargestplace fieldsis
unclear, recordingsmadefromtheventralhippocampus
ofratssuggeststhatfieldsmightexceed10mindiameter
[43];thoughlargerthanatypicalexperimentalroomthis
ismuchsmallerthantherangeofwildratswhichcanbe
hundreds ofmetres [44].
Bycontrasttoplacecells,thespatialactivityofgridcellsis
inherently regular, spanning the available space with
repetitive firing patterns[19]that mayprovide aspatial
metric(thoughsee[45]).Inthemedialentorhinalcortex
medial entorhinalcortex(mEC)gridcells areknownto
exist in functional modules, the cells in each module
havinggrid-likefiringpatternsthatareeffectively
trans-lationsofone another;sharingthesameorientationand
scalebuthavingdifferentoffsetsrelativetothe
environ-ment [19,46–48] (Figure 1b). Modules are distributed
along the dorso-ventral axis of the mEC with those at
moreventrallocationstendingtobeoflargerscalesuch
thatthesizeofthepeaksinthegridfiringpatternandthe
distance betweenthemis increased[19,23,47].Analysis
of thegrid code suggeststhat it provides an extremely
efficientrepresentationofself-location;modulesof
differ-entscalesbehavingsimilarlytotheregistersinaresidue
numbersystemsuchthatcapacityofthenetworkgreatly
exceedsthescaleofthelargestgrid[49,50].Becauseof
these propertiesgridcells arecurrently thoughtto bea
corecomponentoftheneuralsystemresponsibleforpath
integration;theirrepetitivefiringfieldsbeinga
cumulat-iverepresentationofself-motioncues(e.g.[51,52]).Itis
interestingthentonotethatnavigationisnotdissimilarto
the inverseof path integration:the former requires the
calculation of the vector between two allocentric
locations, whilethelatterusesrecentmotion,expressed
as a vector, to update an allocentric representation of
self-location. As such it seems possible that the neural
architecturethatsupportspathintegrationmightalsoplay
arolein navigation.
Indeed,severalauthorshaverecentlyproposedmodelsof
navigation in which grid cells are seen as the central
componentofanetworkabletodeterminetheallocentric
vectorbetweenananimal’scurrentlocationanda
remem-beredgoal[53–55].However,themechanismsemployed
bythemodelsdiffermarkedly,rangingfromaniterative
search for the appropriate vector [53] to a complex
representation of all possible vectors projected into to
thecyclicgridspace[54].Assuch,attheneurallevel,itis
stilltooearlytopredicthowtheactivityofindividualgrid
thepopulationlevelaccessibleto fMRI,itseems
plaus-iblethatmetabolicactivityintheentorhinalcortexshould
correlatewithallocentricspatialparameters.Indeeditis
alreadyknownthatthecoherenceofthedirectionalsignal
associated with grid cells correlates with navigational
performance[56].Furthermore,inlightofthelimitations
imposedonplace cellmodelsofnavigationbythe
irre-gulardistributionofplacefields,itseemsmorelikelythat
activity in the hippocampus will reflect route based
variables.
Neural
representations
of
the
distance
to
the
goal
AnumberofrecentfMRIstudieshaveexaminedwhether
brain activity is correlated with the distance between
landmarksor togoalsduring navigation.During
naviga-tionanumberof spatial parametersrepresentthe
navi-gator’s relationship to the goal (Figure 2a) and these
parameters change over the different key events and
epochsthatcharacterisenavigation(Figure2b).Humans
have been shown to be reasonably good at estimating
Figure2 Goal 1 3 2 3 2 2 3 4 2 3 2 5 &% 1 3 2 4 5
Key Events/Epochs in Navigation Initial Route Planning Travel Periods Path Choice Forced Detour Arrival at the Goal Path to goal Path obstructed
(a)
(b)
University College London
Goal Person Navigating Barrier Path Distance α α Euclidean Distance β β Egocentric Direction θ θ ω ω Allocentric Direction N
Current Opinion in Behavioral Sciences
Spatialgoalparametersandkeyevents/epochsinnavigation.(a)Fourdifferentspatialrelationshipsaredepictedbetweenapersonnavigatingand theirgoal.Pathdistancereferstothedistancealongthepathtotheirgoal(alsoreferredtoasthe‘city-blockdistance’or‘geodesicdistance’).The Euclideandistanceisthedistancealongtheshorteststraightlineconnectingcurrentpositionandthegoal.Egocentricdirectionistheanglebetween person’scurrentfacingdirectionandthedirectionalongtheEuclidean.Theallocentricdirectionisanglebetweenafixedreferencedirectioninthe environment(e.g.North)andtheEuclidean.(b)AmapofpartofUniversityCollegeLondon,withroutebetweenthefirstauthor’soffice(blacktriangle) andthesecondauthor’soffice(goal)shown.Fivekeyevents/epochsareshownalongthejourney.Aforceddetourisillustratedwhereachangeinthe pathisrequired.
parameterssuchasEuclideandistance,pathdistance,and
direction to distant locations, at least in large complex
buildings [57]. Two studies have reported increased
activity in themidto anterior hippocampusatthestart
of navigationwhen route planning was required [8,58].
Such activitymayrelateto theinitialdemands of
plan-ning the route to the goal, however it was not clear
whether this activity was related to the distance to the
goal.ThefirstfMRIstudytoexaminespatialgoalcoding
foundthatactivityintheentorhinalcortexofLondontaxi
drivers was significantly positively correlated with the
Euclidean distance tothe goalduring thenavigationof
avirtualsimulationofLondon,UK[9](Figure3a).This
resultisconsistentwiththeentorhinalcortexcoding an
allocentric vector to thegoal [53–55,59]. Severalrecent
studies haveadopted asimilar approach (Figure 3b–d).
Thesestudiesvarysubstantiallyintermsofthetypesof
environments(e.g. acityregionversusterraindevoidof
landmarks), the amount of prior learning (e.g. 4 years
versus10s),andthetaskrequired(navigatetoa
remem-bered goal versus choosing the path to a visible goal).
Despite these differences all studies have consistently
reportedasignificantrelationshipbetweenhippocampal
activityandgoalproximity.However,lessconsistenthave
beenthesignofthecorrelations(seeFigure3b–d),with
some studies reporting a positive correlation [52] and
othersanegativecorrelation [53,54].
ArecentstudybyHowardetal.[55]providessomeinsight
intotheseapparentlyconflictingresults,andthe
respect-ive rolesthehippocampusand entorhinalcortexduring
thedifferentstages of navigation(shown in Figure2b).
Howardetal.hadsubjectslearn,viaamapandawalking
tour,apreviouslyunfamiliarreal-worldenvironmentand
onthefollowingdaynavigateto goalsin avirtual
simu-lationoftheenvironment(Figure3e).Routesnavigated
were designed such that they separated the Euclidean
distancefromthepathdistancetothegoalandpermitted
brainactivityduringthevariousstagesofnavigationtobe
examined (Figure 2b). While posterior hippocampal
activitywas correlatedwith thepathdistance atseveral
stages of navigation, entorhinal activity was correlated
withthechangeintheEuclideandistancetogoalwhen
initially planningtheroute.Thus,consistentwith some
computationalperspectives,theentorhinalcortexmight
provide information for a goal vector and the
hippo-campus processesthepathto thegoal[53–55,59].
Howard et al. also found that the relationship between
hippocampalactivityandthedistancetothegoaldiffered
dependingontheoperationalstageofnavigation.At
path-choicepointshippocampalactivitywasnegatively
corre-latedwiththedistance(andwithorientation)tothegoal
(i.e. increasingwithgoalproximity),whileduringtravel
periods it waspositivelycorrelated withthe distanceto
the goal (Figure 3e). When the task demands in other
studies reporting activity correlated with distance
(Figure 3a–d)are considereda similar patternemerges.
In tasks involving either purely path decisions [53] or
multipledecisionsinquicksuccessionaboutthedirection
totravel[54],anegativecorrelationbetweenactivityand
distance was observed (Figure 3c,d). Whilst, in studies
involving updating locations viewed [51], or mainly
updating self-location during travel [50], activity was
positively correlated with the distance to the goal
(Figure 3a,b). One possibility is that updating the
dis-tancetoagoalismoredemandingwhenfarfromthegoal,
leadingtoapositivecorrelation.Thiswouldbeconsistent
withstudieslinkinghippocampalactivitytospatial
updat-ingdemands[64–66].Activityincreasingwithproximity
tothegoalatpathchoicepointsmayrelatetoreportsof
hippocampal place cell activity clustered near goals
[32,33–35], whichwould lead to anegative correlation
between distance and activity. More research will be
required to determine how task demands relate to
dis-tancecoding inthehippocampusandentorhinalcortex.
Apotentialpitfallwithstudiesusingcorrelationsbetween
parametric parameters and brain activity is that
uncon-trolledpropertiesofthestimulimightberesponsiblefor
mediating the effects. By including a control condition
Howardetal.revealedthatsimplybeingledtothegoal
was not sufficient to elicit a significant correlation
be-tweenactivityandthedistance.Thus,representing
infor-mation related to the distance to the goal in the
hippocampus and entorhinal cortex appears to require
active goal-directed navigation. An important line of
future enquiry will beto determine whether the
corre-lationsbetweenMTLactivityanddistancearerelatedto
other factorsinvolvedingoal-directed navigation.
Three important factors that mayco-vary with the
dis-tanceto thegoalare:firstlymemorydemands,secondly
thetimerequiredtotraveltothegoalandfinallyreward
associatedwithreachingthegoal.Recallingtherouteto
far away goal locations would arguably make greater
demands onretrievalof theenvironment thanrecalling
the route to close by locations. Thus, it may be that
retrieval demands might underlie the positive
corre-lations observed betweenhippocampal activity and the
distance to thegoal.Ithasbeen arguedthatthe
hippo-campal role in navigation is purely to retrieve stored
knowledge of the environment, not to make the path
calculations [67]. Independently manipulating the
dis-tance from the number of turns and junctions along a
route would help determine whether the hippocampus
processesinformationrelated directlyto thedistanceor
processinformationrelatedtothenumberoffragmentsof
theenvironmentthatconstitutetheroute.Hippocampal
cells have recently been found to code for the time
elapsed during navigation [68] and to modulate their
activity depending on future rewards [69], thus it is
possible that the time required to reach the goal or
Figure3
Start Goal
Plan View
Whole Journey Whole Journey
City region learned an average of 4 years prior to testing
Navigation to a goal learned 10 sec prior in an environment with no landmarks
Path Choice
BOLD signal Δ Euclidean
Distance
City region learned one day prior to testing VR room learned minutes prior to tesing
BOLD signal Path Distance Viewing sequentially presented landmark images
BOLD signal Distance
Between Landmarks
City learned an average of 16 years prior to testing
(a)
(b)
(c)
(d)
(e)
BOLD signal Euclidean Distance
BOLD signal
Distance
BOLD signal Distance
Path Choice and Travel Periods
BOLD signal
Path Distance X Goal Direction
Environment / Task Correlations between fMRI signal and spatial goal parameters
Entorhinal Cortex Entorhinal Cortex Hippocampus Hippocampus L R Hippocampus Hippocampus Hippocampus
BOLD signal Egocentric Goal Direction
BOLD signal Egocentric Goal Direction Travel Periods
Travel Periods Initial Route Planning
Path Choice
Parietal Cortex
Parietal Cortex
BOLD signal Δ Path Distance Hippocampus
Spiers and Maguire, 2007
Morgan et al., 2011 1 1 1 2 3 4 5 2 3 4 5 6 7 8 10 9 Sheril et al., 2013 Viard et al., 2011 Howard et al., 2014 Forced Detour 50 metres
Egocentric Goal Direction Euclidean Distance Path Distance Current Location at 150 sec New Goal Event Route Goal Location 2 1 3 4 5 10 seconds 8 seconds 2 seconds Time Map Presentation Delay
First person perspective
hippocampalactivityanddistance.Futureneuroimaging
studies which vary reward, time and distance, will be
helpfulinteasingapartthesepossibilities,aswillresearch
directlytestingwhetherneuronalfiringpatternsare
cor-related withspatialgoalparameters.
Animportantrecentsingleunitrecordingstudyexplored
howhippocampalplacecellactivityrelatedtothe
trajec-torytothefuturegoalduringnavigationepochs.Pfeiffer
and Foster [70] recorded CA1 place cells while rats
foraged for rewardsin anopen fieldenvironment.After
foraging for, and finding, a reward in the arena rats
returned to arewarded ‘home’ location that was stable
withinaday,butchangeddaytoday.PfeifferandFoster
foundthatbeforetravellingtothegoal,duringensemble
populationspikingeventsinCA1,thebriefactivationof
placecellscodinglocationsbetweentheratanditsfuture
goaloccurred.Theactivationwasnotafaithful‘read-out’
oftheexactfuturepath,butappearedtoencompassarange
ofpossibletrajectorypositionsfallingbetweentheratand
its future goal. Although notquantified in thestudy, it
appearsthatthelongerthedistancethegreaterthenumber
ofcellsactivatedinthe populationsspikingevents.This
wouldpotentiallyprovideanexplanationfor why
hippo-campalactivitymaybegreaterwhenthenavigatorisfar
fromtheirgoal[61].However,suchamechanismcannot
explainwhyactivityincreaseswithproximitytothegoal
whenchoosingthepath(Figure3).Thusitislikelythat
multiple mechanisms operate in the hippocampus to
codeinformationaboutspatialgoals.
Neural
representations
of
the
direction
to
the
goal
Whileemergingdataimplicatestheentorhinalregionin
codingtheEuclideandistancealongavectortothegoal
[50,55],itisnotyetclearwhetherentorhinalgridcells,or
conjunctivegridcellsunderliethisphenomenon.Models
predict that the allocentric direction to the goal
(Figure2a)isinitiallycomputedinmedialtemporallobe
structuresandsubsequently convertedtotheegocentric
directiontoguidebodymovementthroughspace[53,71].
Consistent with this two fMRI studies have reported
activity patterns in posterior parietal cortex associated
with the egocentric direction to the goal during travel
periods ([50,55]; Figure 3a,e). Evidence for allocentric
goaldirectioncodinghasyettobereported,andthusits
existenceiscurrently onlyatheoreticalprediction.
Conclusion
Recentcomputational models, fMRI,
electrophysiologi-calstudieshavebeguntoshedlightonhowthebrainmay
encodethespatialrelationshiptothegoalduring
naviga-tion.Currentevidenceimplicatestheentorhinalcortexin
codingthedistancealongavectortothegoal,the
hippo-campus representing thepath to thegoal and posterior
parietalcortexcodingtheegocentricdirectiontothegoal.
Howhippocampal activityrelatesto thedistanceto the
goal, appears to depend upon the operational stage of
navigation, whetherthenavigatoristravelling,choosing
thepath,orplanningtheroute.Futureresearch
integrat-ing rodent electrophysiology and neuroimaging data to
test modelpredictionswill beimportant toadvance our
understanding of theneuralsystemssupporting
naviga-tionalguidance.
Conflict
of
Interest
Nothingdeclared
Acknowledgements
ThisworkwasfundedbyaWellcomeTrustgrant(094850/Z/10/Z)and JamesSMcDonnellScholarAwardtoHJSandaSirHenryDaleFellowship jointlyfundedbytheWellcomeTrustandRoyalSocietytoCB.
References
and
recommended
reading
Papersofparticularinterest,publishedwithintheperiodofreview, havebeenhighlightedas:
ofspecialinterest ofoutstandinginterest
1. PackardMG,McGaughJL:Inactivationofhippocampusor caudatenucleuswithlidocainedifferentiallyaffects expressionofplaceandresponselearning.NeurobiolLearn Mem1996,65:65-72.
2. BerkeJD,BreckJT,EichenbaumH:Striatalversushippocampal representationsduringwin-staymazeperformance.J Neurophysiol2009,101:1575-1587.
3. VanderMeerM,Kurth-NelsonZ,RedishAD:Information processingindecision-makingsystems.Neuroscientist2012, 18:342-359.
4. ScovilleWB,MilnerB:Lossofrecentmemoryafterbilateral hippocampallesions.JNeurolNeurosurgPsychiatry1959, 20:11-21.
(Figure3Legend)fMRIstudiesreportingbrainactivitycorrelatedwithspatialgoalparameters.Ontheleftareshownexamplesofthestimuliand environmentsusedinthestudies.Ontherightareshownstatisticalparametricmaps(SPMs)overlaidonstructuralimagesshowingregionswith activitycorrelatedthespatialparameters.NexttoeachSPMisaschematicillustrationofwhetherthecorrelationwaspositiveornegativeandthe event/epoch(seeFigure2b)examinedlistedabovetheSPM.(a)AmapwithoneofthesevenroutesdriveninthesimulationofLondon(UK)andimage fromwiththesimulationatTrafalgarSquare(figureadaptedfrom[9]).(b)Fiveexamplesfromthelargersetoflandmarkssequentiallyshownto subjectsduringscanning,belowisamapoftheUniversityofPennsylvaniacampuswithlandmarklocationsmarked(figureadaptedfrom[60]).(c) Left:exampleofastimulusviewedduringtesting,middle:aplanviewoftheroom,notshowntosubjects.Subjectshadtochoosetheshortestrouteto thegoalfromarangeofdifferentstartingpointsalongthebackwalloftheroom(figureadaptedfrom[62]).(d)Subjectsviewaplanviewofamap indicatingtheircurrentlocationandgoallocationinaVRworldcomposedoftessellatingobstructiveblackcylindersonagreyplane.Aftera10sdelay withfixationsubjectsnavigatedtowheretheythoughtthegoalwaslocated(figureadaptedfrom[63]).(e)Amapwithoneofthe10routesnavigatedin theSohoregionofLondon(UK)(figureadaptedfrom[61]).ThepathdistanceandEuclideandistancetothegoalareshownillustratedatatimepoint 150sintotheroute.Thefilmstripshowashortsegmentofthefootageusedtosimulatetheenvironment.IntheillustrativeplotsD=thechangeinthe parameter.Largechangesintheparametersoccurredwhenanewgoalwaspresentedandinitialrouteplanningwasneedorwhenaforceddetour occurred.
5. MorrisRGM,GarrudP,RawlinsJNP,O’KeefeJ:Placenavigation impairedinratswithhippocampallesions.Nature1982, 297:681-683.
6. SpiersHJ,BurgessN,MaguireEA,BaxendaleSA,HartleyT, ThompsonPJ,O’KeefeJ:Unilateraltemporallobectomy patientsshowlateralizedtopographicalandepisodicmemory deficitsinavirtualtown.Brain2001,124:2476-2489. 7. HartleyT,MaguireEA,SpiersHJ,BurgessN:Thewell-wornroute
andthepathlesstraveled:distinctneuralbasesofroute followingandwayfindinginhumans.Neuron2003,37:877-888. 8. SpiersHJ,MaguireEA:Thoughts,behaviour,andbrain
dynamicsduringnavigationintherealworld.Neuroimage2006, 31:1826-1840.
9.
SpiersHJ,MaguireEA:Anavigationalguidancesysteminthe humanbrain.Hippocampus2007,17:618-626.
Firststudytoexamine thedistanceanddirection to thegoalduring navigationwithfMRI.
10. VannSD,AggletonJP,MaguireEA:Whatdoestheretrosplenial cortexdo? NatRevNeurosci2009,10:792-802.
11. EpsteinRA:Parahippocampalandretrosplenialcontributions tohumanspatialnavigation.TrendsCognSci2008,12:388-396. 12. PoucetB,Lenck-SantiniPP,HokV,SaveE,BanquetJP,
GaussierP,MullerRU:Spatialnavigationandhippocampal placecellfiring:theproblemofgoalencoding.RevNeurosci 2004,15:89-107.
13. SpiersHJ:Keepingthegoalinmind:prefrontalcontributionsto spatialnavigation.Neuropsychologia2008,46:2106-2108. 14. RochefortC,AraboA,AndreM,PoucetB,SaveE,Rondi-ReigL:
Cerebellumshapeshippocampalspatialcode.Science2011, 334:385-389.
15. BarryC,BurgessN:Neuralmechanismsofself-location.Curr Biol2014,24:R330-R339.
16. O’KeefeJ,DostrovskyJ:Thehippocampusasaspatialmap. Preliminaryevidencefromunitactivityinthefreely-movingrat. BrainRes1971,34:171-175.
17. MullerRU,KubieJL,Ranck JBJr:Spatialfiringpatternsof hippocampalcomplex-spikecellsinafixedenvironment. JNeurosci1987,7:1935-1950.
18. BostockE,MullerRU,KubieJL:Experience-dependent modificationsofhippocampalplacecellfiring.Hippocampus 1991,1:193-206.
19. HaftingT,FyhnM,MoldenS,MoserM-B,MoserEI:
Microstructureofaspatialmapintheentorhinalcortex.Nature 2005,436:801-806.
20. BoccaraCN,SargoliniF,ThoresenVH,SolstadT,WitterMP, MoserEI,MoserM-B:Gridcellsinpre-andparasubiculum.Nat Neurosci2010,13:987-994.
21. TaubeJS,MullerRU,RanckJB:Head-directioncellsrecorded fromthepostsubiculuminfreelymovingrats.II.Effectsof environmentalmanipulations.JNeurosci1990,10:436-447. 22. SolstadT,BoccaraCN,KropffE,MoserM-BB,MoserEI:
Representationofgeometricbordersintheentorhinalcortex. Science2008,322:1865-1868.
23. SargoliniF,FyhnM,HaftingT,McNaughtonBL,WitterMP, MoserM,MoserEI:Conjunctiverepresentationofposition, direction,andvelocityinentorhinalcortex.Science2006, 312:758-762.
24. ValerioS,TaubeJS:Pathintegration:howtheheaddirection signalmaintainsandcorrectsspatialorientation.NatNeurosci 2012,15:1445-1453.
25. VanderMeerMAA,RichmondZ,BragaRM,WoodER, DudchenkoPA:Evidencefortheuseofaninternalsenseof directioninhoming.BehavNeurosci2010,124:164-169. 26. O’KeefeJ,SpeakmanA:Singleunitactivityintherat
hippocampusduringaspatialmemorytask.ExpBrainRes 1987,68:1-27.
27. KomorowskiRW,MannsJR,EichenbaumH:Robustconjunctive item-placecodingbyhippocampalneuronsparallelslearing whathappenswhere.JNeurosci2009,29:9918-9929. 28. AingeJA,TamosiunaiteM,WoergoetterF,DudchenkoPA:
HippocampalCA1placecellsencodeintendeddestination onamazewithmultiplechoicepoints.JNeurosci2007, 27:9769-9779.
29. WoodER,DudchenkoPA,RobitsekRJ,EichenbaumH: Hippocampalneuronsencodeinformationaboutdifferent typesofmemoryepisodesoccurringinthesamelocation. Neuron2000,27:623-633.
30. O’KeefeJ,ConwayDH:Hippocampalplaceunitsinthefreely movingrat:whytheyfirewheretheyfire.ExpBrainRes1978, 31:573-590.
31. VanderMeerMAA,JohnsonA,Schmitzer-TorbertNC,RedishAD: Tripledissociationofinformationprocessingindorsal striatum,ventralstriatum,andhippocampusonalearned spatialdecisiontask.Neuron2010,67:25-32.
32.
DupretD,O’NeillJ,Pleydell-BouverieB,CsicsvariJ:The reorganizationandreactivationofhippocampalmapspredict spatialmemoryperformance.NatNeurosci2010,13:995-1002. ProvidesevidenceforCA1,butnotCA3re-mappinginrelationto goal-locationsduringupdatinggoallocations.
33. FyhnM,MoldenS,WitterMP,MoserEI,MoserM:Spatial representationintheentorhinalcortex.Science2004, 305:1258-1264.
34. HollupSA,MoldenS,DonnettJG,MoserMB,MoserEI: Accumulationofhippocampalplacefieldsatthegoal locationinanannularwatermazetask.JNeurosci2001, 21:1635-1644.
35. HokV,Lenck-SantiniP-P,RouxS,SaveE,MullerRU,PoucetB: Goal-relatedactivityinhippocampalplacecells.JNeurosci 2007,27:472-482.
36. EkstromAD,KahanaMJ,CaplanJB,FieldsTA,IshamEA, NewmanEL,FriedI,DaleJ,KandilF,NishidaS:Cellular networksunderlyinghumanspatialnavigation.Nature2003, 425:184-188.
37. BurtonBG,HokV,SaveE,PoucetB:Lesionoftheventraland intermediatehippocampusabolishesanticipatoryactivityin themedialprefrontalcortexoftherat.BehavBrainRes2009, 199:222-234.
38. BurgessN,O’KeefeJ:Neuronalcomputationsunderlyingthe firingofplacecellsandtheirroleinnavigation.Hippocampus 1996,6:749-762.
39. BurgessN,JacksonA,HartleyT,O’KeefeJ:Predictionsderived frommodellingthehippocampalroleinnavigation.BiolCybern 2000,83:301-312.
40. FosterDJ,MorrisRG,DayanP:Amodelofhippocampally dependentnavigation,usingthetemporaldifferencelearning rule.Hippocampus2000,10:1-16.
41. MullerRU,SteadM,PachJ:Thehippocampusasacognitive graph.JGenPhysiol1996,107:663-694.
42. O’KeefeJ,NadelL:TheHippocampusasaCognitiveMap. Oxford:OxfordUniversityPress;1978,.
43. KjelstrupKB,SoldstadT,BrunVH,HaftingT,LeutgebS, WitterMP,MoserEI,MoserM-B,SolstadT:Finitescaleof spatialrepresentationinthehippocampus.Science2008, 321:140-143.
44. RussellJC,TownsDR,AndersonSH,CloutMN:Interceptingthe firstratashore.Nature2005,437:1107.
45. DerdikmanD,WhitlockJR,TsaoA,FyhnM,HaftingT,MoserM-B, MoserEI:Fragmentationofgridcellmapsina
multicompartmentenvironment.NatNeurosci2009, 12:1325-1332.
46. BarryC,HaymanR,BurgessN,JefferyKJ: Experience-dependentrescalingofentorhinalgrids.NatNeurosci2007, 10:682-684.
47. StensolaH,StensolaT,SolstadT,FrølandK,MoserM-B, MoserEI:Theentorhinalgridmapisdiscretized.Nature2012, 492:72-78.
48. YoonK,BuiceMA,BarryC,HaymanR,BurgessN,FieteIR: Specificevidenceoflow-dimensionalcontinuousattractor dynamicsingridcells.NatNeurosci2013,16:1077-1084. 49. FieteIR,BurakY,BrookingsT:Whatgridcellsconveyaboutrat
location.JNeurosci2008,28:6858-6871.
50. TowseBW,BarryC,BushD,BurgessN:Optimalconfigurations ofspatialscaleforgridcellfiringundernoiseanduncertainty. PhilosTransRSocLondBBiolSci2014,369.
51. BurgessN,BarryC,O’KeefeJ:Anoscillatoryinterference modelofgridcellfiring.Hippocampus2007,17:801-812. 52. McNaughtonBL,BattagliaFP,JensenO,MoserEI,MoserMB:
Pathintegrationandtheneuralbasisofthe‘‘cognitivemap’’. NatRevNeurosci2006,7:663-678.
53. ErdemUM,HasselmoM:Agoal-directedspatialnavigation modelusingforwardtrajectoryplanningbasedongridcells. EurJNeurosci2012,35:916-931.
54. KubieJL,FentonAA:Linearlook-aheadinconjunctivecells:an entorhinalmechanismforvector-basednavigation.Front NeuralCircuits2012http://dx.doi.org/10.3389/fncir.2012.00020.
55. BarryC,BushD,BurgessN:Howtousegridcellsfor high-precisionlarge-scalevectornavigation.SfNAbstr2014,848. 56. DoellerCF,BarryC,BurgessN:Evidenceforgridcellsina
humanmemorynetwork.Nature2010,463:657-661.
57. ThorndykePW,Hayes-RothB:Differencesinspatialknowledge acquiredfrommapsandnavigation.CognPsychol1982, 14:560-589.
58. XuJ,EvensmoenHR,LehnH,PintzkaCWS,Ha˚bergAK: Persistentposteriorandtransientanteriormedial temporallobeactivityduringnavigation.Neuroimage2010, 52:1654-1666.
59. BarryC,BushD:FromAtoZ:apotentialroleforgridcellsin spatialnavigation.NeuralSystCircuits2012,2:6.
60.
MorganLK,MacevoySP,AguirreGK,EpsteinRA:Distances betweenreal-worldlocationsarerepresentedinthehuman hippocampus.JNeurosci2011,31:1238-1245.
Providedkeyevidencefromanon-navigationtaskthatsimplyviewing landmarkscanactivatedistancerepresentations.
61.
HowardLR,JavadiAH,YuY,MillRD,MorrisonLC,KnightR, LoftusMM,StaskuteL,SpiersHJ:Thehippocampusand entorhinalcortexencodethepathandEuclideandistancesto goalsduringnavigation.CurrBiol2014,24:1331-1340. Providesa breakdown ofthedifferentstages ofnavigation andthe correlationswithdistanceandegocentricdirectiontothegoalatthese differenttimeperiods.
62. ViardA,DoellerCF,HartleyT,BirdCM,BurgessN:Anterior hippocampusandgoal-directedspatialdecisionmaking. JNeurosci2011,31:4613-4621.
63.
SherrillKR,ErdemUM,RossRS,BrownTI,HasselmoME, SternCE:Hippocampusandretrosplenialcortexcombinepath integrationsignalsforsuccessfulnavigation.JNeurosci2013, 33:19304-19313.
Reportsevidencethatthehippocampusiscorrelatedwiththedistanceto thegoalduringnavigationofalandmark-freeenvironment.
64. HowardLR,KumaranD,O´ lafsdo´ttirHF,SpiersHJ:Double dissociationbetweenhippocampalandparahippocampal responsestoobject-backgroundcontextandscenenovelty. JNeurosci2011,31:5253-5261.
65. WolbersT,WienerJM,MallotHA,Bu¨chelC:Differential recruitmentofthehippocampus,medialprefrontalcortex,and thehumanmotioncomplexduringpathintegrationinhumans. JNeurosci2007,27:9408-9416.
66. KumaranD,MaguireEA:Whichcomputationalmechanisms operateinthehippocampusduringnoveltydetection? Hippocampus2007,748:735-748.
67. EichenbaumH,CohenNJ:Canwereconcilethedeclarative memoryandspatialnavigationviewsonhippocampal function? Neuron2014,83:764-770.
68. KrausBJ,RobinsonRJ,WhiteJA,EichenbaumH,HasselmoME: Hippocampal‘‘timecells’’:timeversuspathintegration. Neuron2013,78:1090-1101.
69. LeeH,GhimJ-W,KimH,LeeD,JungM:Hippocampalneural correlatesforvaluesofexperiencedevents.JNeurosci2012, 32:15053-15065.
70.
PfeifferBE,FosterDJ:Hippocampalplace-cellsequences depictfuturepathstorememberedgoals.Nature2013, 497:74-81.
Indicatesthatplacecellsencodinglocationsalongpotentialpathstoa goalare‘pre-activated’beforenavigation.
71. ByrneP,BeckerS,BurgessN:Rememberingthepastand imaginingthefuture:aneuralmodelofspatialmemoryand imagery.PsycholRev2007,114:340-375.
