invited review anatomical loops and their electrical ... · correspondence: david kleinfeld,...

Correspondence David Kleinfeld Department of Physics University of California 9500 Gilman Drive La Jolla CA 92093plusmn 0319 USATel (619) 822 plusmn 0342 Fax (619) 534 plusmn 7697 E-mail dkphysicsucsdedu

Invited Review

Anatomical loops and their electrical dynamics in relation to

whisking by rat

DAVID KLEINFELD RUNE W BERG and SEAN M Orsquo CONNOR

Department of Physics University of California La Jolla CA 92093 USA

AbstractAn accumulation of anatomical behavioral and electrophysiological evidence allows us to identify the neuronal circuitrythat is involved with vibrissa-mediated sensation and the control of rhythmic vibrissa movement Anatomical evidence pointsto a multiplicity of closed sensorimotor loops while electrophysiological data delineate the flow of electrical signals in thesepathways These loops process sensory input from the vibrissae and send projections to direct vibrissa movement startingat the level of the hindbrain and proceeding toward loops that involve multiple structures in the forebrain The nature of thevibrissa-related electrical signals in behaving animals has been studied extensively at the level of neocortical loops Two typesof spike signal are observed that serve as a reference of vibrissa motion a fast signal that correlates with the relative phaseof the vibrissae within a whisk cycle and a slow signal that correlates with the amplitude and possibly the set-point of thevibrissae during a whisk Both signals are observed in vibrissa primary sensory (S1) cortex and in some cases they aresufficiently robust to allow vibrissa position to be accurately estimated from the spike train of a single neuron Unlike thecase for S1 cortex only the slow signal has been observed in vibrissa primary motor (M1) cortex The control capabilitiesof M1 cortex were estimated from experiments with anesthetized animals in which progressive areas along the vibrissa motorbranch were microstimulated with rhythmically applied currents The motion of the vibrissae followed stimulation of M1cortex only for rates that were well below the frequency of rhythmic whisking in contrast the vibrissae followed stimulationof the facial nucleus whose cells directly drive the vibrissae for rates above that of whisking In toto the evidence impliesthat there is fast signaling from the facial nucleus through the mystacial pad and the vibrissae and up through sensorycortex but only slow signaling at the level of the motor cortex and down through the superior colliculus to the facial nucleusThe transformation from fast sensory signals to slow motor control is an unresolved issue On the other hand there is acandidate scheme to understand how the fast reference of vibrissa motion in the whisk cycle may be used to decode the angleof the vibrissae upon their contact with an object We discuss a circuit in which servo mechanisms are used to determine theangle of contact relative to the preferred phase of the fast reference signals Support for this scheme comes from results withanesthetized animals on the frequency and phase entrainment of intrinsic neuronal oscillators in S1 cortex A predictionbased on this scheme is that the output from a decoder circuit is maximal when the angle of contact differs from the preferredphase of a fast regerence signal In contrast for correlation-based schemes the output is maximal when the angle of contactequals the preferred phase

Introduction

Spatial navigation requires active sensory as well asmotor processes Animals must extract meaningfrom the sensory information they amass throughtheir receptors as they search and locomote Afundamental question in studies of sensory percep-tion is how the blur of sensory input is converted bythe nervous systems into a stable perception Herewe review the question of active sensation in thecontext of tactile localization of objects accomplishedby the exploratory whisking movement of vibrissae inthe rat Our goal is to focus the quest for thealgorithm that allows rats to extract a stable image ofthe world from the input generated by the activemovement of their vibrissae This experimentalapproach was laid down by Vincent (1912) and was

followed only decades later by Welkerrsquos (1964)study and more contemporary studies that involvedtrained animals (Hutson and Masterton 1986Guic-Robles et al 19891992 Carvell and Simons1990 1995 Barneoud et al 1991 Bermejo et al1996) and electrophysiological recording (Fee et al1997)

Our emphasis is on the structure and function ofclosed-loop pathways in the sensorimotor vibrissasystem that are expected to play an essential role asrats use their vibrissae to search their environment(Fig 1a) As such we have organized this review toreflect the following issues (i) the mechanics ofwhisking (ii) the potential involvement of multipleclosed neuronal loops in the sensorimotor pathway(iii) the nature of exploratory whisking vs tremormotion in sessile animals (iv) neuronal correlates of

0899 plusmn 022099020069 plusmn 20 1999 Taylor amp Francis Ltd

Somatosensory amp Motor Research 1999 16(2) 69 plusmn 88

whisking at the level of sensory and motor cortex (v)the nature of vibrissa control along the motorpathway and (vi) mechanisms for computing the

position of the vibrissae upon their contact with anobject

Mechanics of whisking

The face of the rat has a muscular thickeningdenoted the mystacial pad that contains upwardsof 40 long hairs referred to as vibrissae that arenominally arranged in a Manhattan-like grid withfive rows Each vibrissa sits in a follicle that con-tains sensory nerve endings that originate from abranch of the trigeminal nerve (infraorbital branchof the 5th cranial nerve) (Dorfl 1985 Rice et al1986) The follicle is attached to the mystacial padnear the surface of the skin This attachment servesas a pivot and the follicle is propelled by contrac-tion of a muscular sling that is innervated by abranch of the facial nerve (7th cranial nerve) (Fig1b) (Dorfl 1985 Wineski 1985 Rice et al1994) One feature of this arrangement is thatprotraction is active while retraction is passiveThus the frequency of whisking is ultimately lim-ited by the viscoelastic properties of the mystacialpad A second feature of this arrangement is thatactivation of the sensory afferents is expected todepend primarily on a change in the position ofthe vibrissa relative to that of the follicle as mayoccur upon contact of the vibrissae with an exter-nal object or during periods of rapid accelerationof the vibrissae in the absence of contact

The motion of the vibrissae is governed bymuscles that move the entire mystacial pad as wellas the muscles that control each follicle (Wineski1985) Despite the potential for individual vibris-sae to move independent of each other the vibris-sae are observed to move largely as a single unitduring a multitude of exploratory behaviors (Vin-cent 1912 Wineski 1983) Further the vibrissaemove with bilateral symmetry (Fig 1a) Thuswhisking may be described solely in terms of singleangle as a function of time at least to the extentthat translational motion of the mystacial pad canbe ignored The change in position is described interms of a rapidly changing phase within eachwhisk cycle and a slowly changing set-point andmaximum amplitude (Fig 1c) We will show laterthat this description allows the position of thevibrissae to be measured via the electromyogram(EMG) at a single location in the mystacial pad

A final issue concerns the lack of proprioceptivefeedback of vibrissa position In skeletal joints theextent of muscle contraction is coded by the innerva-tion of specialized muscle fibers known as spindlefibers that provide feedback on the actual motion ofthe muscles as part of a reflex loop However there isno evidence for spindle fibers in facial musculature(Bowden and Mahran 1956) Consistent with thisgeneral result unpublished anatomical studies (F LRice personal communication) have confirmed theabsence of spindle fibers throughout the musculature

70 D Kleinfeld et al

FIGURE 1 The nature and control of exploratory whiskingin the rat (a) Consecutive video frames of the head of a ratas it produces large exploratory whisks in search of a foodtube The vibrissae were highlighted by strob illuminationThe time between frames is 33 ms (b) The sling-likearrangement of the musculature surrounding each folliclein the mystacial pad The lower structure is one of twosensory nerves (adapted from Dorfl (1982)) (c) Diagramof the cyclic motion of the vibrissae during exploratorywhisking The animal whisks with frequency n D The set-point and maximum amplitude of the rhythmic whiskingvary slowly on the timescale of each whisk cycle

of the mystacial pad Thus rats must infer theposition of their vibrissae through spikes generatedby the sensory afferents a scheme known as periph-eral reafference or by neuronal reference signals thatare generated at the central level a scheme known ascorollary discharge

Anatomical loops in vibrissa sensation and

control

A fruitful way to study the gross connectivity withinthe vibrissa sensorimotor network is in terms ofclosed loops The ordf lowestordm of these loops involvesonly hindbrain structures and is confined to theipsilateral side of the brain Higher-order loopsinvolve connections that cross the midline culminat-ing with loops that involve multiple thalamic andneocortical areas By tracing through closed path-ways that involve the vibrissae we hope to illustratethe relevant anatomy while keeping a focus on thepossible neuronal computations and behavioral func-tions that are enabled by this closed-loop systemNote that for the benefit of readability the referencesthat established the connectivity among different lociare listed in the figure captions1

Hindbrain loop

The most compact sensorimotor loop involves inputthat is relayed by the trigeminal ganglion to trigemi-nal nuclei that in turn project to the facial motornucleus which drives the vibrissae (Fig 2) Projec-tions from the neurons of the trigeminal ganglionhave peripheral branches that innervate the vibrissafollicles and have central branches that project totrigeminal nuclei which include the principal sen-sory nucleus and the three spinal nuclei denotedoralis interpolaris and caudalis these projectionsform several somatotopic representations of theipsilateral vibrissae The trigeminal nuclei project tothe lateral subnucleus of the ipsilateral facial nucleusthis subnucleus sends motor output to the mystacialpad to complete the loop The observed pattern ofconnectivity suggests that the trigeminal nuclei mayexert feedback control on whisking However thereare presently no data that show direct connectionsbetween vibrissa trigeminal afferents and the facialmotor neurons that drive the vibrissae

Of critical importance whisking occurs in theabsence of sensory feedback Welker (1964) observedbilateral whisking in animals with a unilateral lesionof the trigeminal nerve and more recently Zieglerreported (personal communication) highly coherentbilateral whisking in animals with bilateral lesionsThese data imply that a yet undiscovered centralpattern generator2 drives rhythmic whisking ahypothesis first suggested in this context by Carvell etal (1991)

Midbrain loop

A higher-level loop incorporates the superior collicu-lus and includes connections that cross the midline(Fig 3) The superior colliculus is a laminarmidbrain structure with each layer nominally devo-ted to integrating sensory and motor informationrelevant to a particular sensory modality (Stein et al1975) In the rat the intermediate and deep layers ofthe colliculus appear to be devoted to somaticsensorimotor processing with the more rostral andlateral areas responding to vibrissa input (Huerta etal 1983 Isokawa-Akesson and Komisaruk 1987Miyashita et al 1994) The middle and deep layersof the rostral plusmn lateral aspects of the superior collicu-lus receive vibrissa-related inputs from the con-tralateral trigeminal nuclei and descending afferentsfrom the superior colliculus project to the lateralsubnucleus of the contralateral facial nucleus Anadditional input that converges to the same laminae

Sensorimotor loops and the whisk cycle 71

FIGURE 2 Hindbrain-level sensorimotor loop (vibrissae

reg trigeminal ganglion ) The vibrissae are innervated bytwo kinds of sensory afferents that originate from theinfraorbital nerve (Vincent 1913 Dorfl 1985 Rice et al1986) (trigeminal ganglion reg trigeminal nuclei)Sensory input from the trigeminal ganglion enters thehindbrain at the trigeminal nuclei consisting of theprincipal sensory nucleus (PrV) and the spinal trigeminalnuclei denoted oralis (SpVO) interpolaris (SpVI) andcaudalis (SpVC) (Cajal 1911 Torvik 1956 Clarke andBowsher 1962) The PrV and SpVI nuclei and themagnocellular portion of SpVC have somatotopic maps ofthe vibrissae (ordf barrelettesordm ) (Ma and Woolsey 1984)SpVO also receives sensory input from the vibrissae yetdoes not contain a map (Belford and Killackey 1979a b)Lastly there is high internuclear connectivity especiallyamong SpVC and SpVO (Jacquin et al 1990a) (trigemi-nal nuclei reg facial nucleus) The facial nucleus containsfive subnuclei of which the lateral subnucleus is involvedin vibrissa control (Papez 1927 Martin and Lodge1977) Vibrissa areas of the trigeminal nuclei SpVC PrVand SpVI connect to the lateral subnucleus primarilythrough ipsilateral projections The dominant projectionappears to arise from the magnocellular portion of SpVC(Erzurumlu and Killackey 1979 Isokawa-Akesson andKomisaruk 1987) (facial nucleus reg vibrissae) Thefacial nucleus sends projections to the papillary musclessurrounding each vibrissa (Arvidsson 1982 Dorfl 19821985 Rice and Arvidsson 1991) The lateral subnucleusof the facial nucleus contains a somatotopic map of thevibrissae (Martin and Lodge 1977)

arises from ipsilateral vibrissa M1 cortex Yet thecomputations that the colliculus performs on theconfluence of vibrissa sensory inputs and motorcommands are presently unknown

Cerebellar loops that involve trigeminal nuclei and the

colliculus

The pontine-cerebellar system appears to function asa hindbrain-level intermediary in a loop that involvesindirect connections between the contralateral tri-geminal nuclei and the ipsilateral superior colliculus(Fig 4) The trigeminal nuclei project to both thepons and the inferior olive which in turn directlyproject to the cerebellum similar inputs whichproject to the same crura in cerebellum arise fromthe intermediate and deep layers of the superiorcolliculus The cerebellar Purkinje cells synapse onthe cerebellar nuclei and this provides outputprojections to superior colliculus to complete theloop

Beyond the issue of loops that directly involve thevibrissae it has been proposed that the embeddedloop between the pontine-cerebellar system and thesuperior colliculus (Fig 4) may function as a rhythmicpattern generator (Westby et al 1993)3 This patterngenerator would not depend on the integrity of thetrigeminal sensory input and thus could function asthe central pattern generator that drives rhythmicvibrissa motion by the facial nucleus

Thalamic forebrain loop

Multiple structures in ventral and dorsal thalamusreceive input from the trigeminal nuclei Only one ofthese zona incerta in the ventral thalamus projectsdirectly back to the superior colliculus (Fig 5)where it forms inhibitory connections with thesuperior colliculus Thus the zona incerta appears tofunction as a forebrain-level intermediary in a loopthat involves the trigeminal nuclei and thecolliculus

Cortical forebrain loops

These high-level loops are formed by projections thatinvolve multiple thalamic nuclei and cortical areasThe thalamocortical branch of the sensorimotorpathway involves projections from the trigeminal


FIGURE 3 Midbrain-level sensorimotor loop (trigeminalnuclei reg superior colliculus) The trigeminal nucleiproject to vibrissa somatotopic areas of the superiorcolliculus (Drager and Hubel 1976 Killackey and Erzur-umlu 1981 Huerta et al 1983 Steindler 1985 Bruce etal 1987 Jacquin et al 1989 Benett-Clarke et al 1992)The connection from interpolaris appears to be thestrongest (Killackey and Erzurumlu 1981 Huerta et al1983 Jacquin et al 1989) while that from caudalis isproblematic (Killackey and Erzurumlu 1981) All connec-tions terminate in the intermediate and deep layers andtend to occur in the lateral and rostral aspects of thecolliculus (Huerta et al 1983) The projections from thetrigeminal ganglia to the colliculus are likely to becollaterals of projections to the thalamus (Mantle-St Johnand Tracey 1987 Benett-Clarke et al 1992) (superiorcolliculus reg facial nucleus) The intermediate and deeplayers of the colliculus project to the lateral subnucleus ofthe facial nerve nucleus (Isokawa-Akesson and Komisaruk1987 Miyashita et al 1994 Miyashita and Shigemi1995)

FIGURE 4 Midbrain-level sensorimotor loops includingprojections with cerebellar nuclei (trigeminal nuclei regcerebellum) The trigeminal nuclei provide vibrissa sen-sory input to the cerebellum via two paths interpolaris andcaudalis project via the inferior olive climbing fibers(Huerta et al 1983 Jacquin et al 1989) and principalisinterpolaris and caudalis project via pontine mossy fibers(Smith 1973 Watson and Switzer 1978 Huerta et al1983 Swenson et al 1984 Steindler 1985 Mantle-StJohn and Tracey 1987) Projections from the trigeminalnuclei to the inferior olive overlap those from the olive tothe cerebellum (Huerta et al 1983) the target areas in thecerebellum include crura I and II (Watson and Switzer1978 Huerta et al 1983) and the paramedian lobule anduvula (Watson and Switzer 1978) all areas with facialreceptive fields (superior colliculus reg cerebellum)The colliculus sends projections to the cerebellar cortexincluding target areas crura I and II through both theinferior olive and the pons (Kassel 1980) (cerebellum regsuperior colliculus) The deep cerebellar nuclei send aprojection to the colliculus (Lee et al 1989 Westby et al1993 1994) which forms a ordf colliculus reg cerebellum regcolliculusordm loop

nuclei to thalamic nuclei from thalamus to sensoryareas and then motor areas in cortex and frommotor cortex down to the superior colliculus tocomplete a loop (Fig 6) Two pathways dominate thesynaptic input of vibrissa sensory information tocortex (see Diamond (1995) and Keller (1995) forreviews of corticothalamic connectivity) The princi-pal trigeminal nucleus projects to the ventral poster-omedial nucleus of dorsal thalamus (lemniscal path)which in turn has major projections to vibrissa S1cortex and minor projections to secondary sensorycortex The spinal trigeminal nuclei project to theposterior nucleus of dorsal thalamus (paralemniscalpath) which in turn projects to vibrissa primary andsecondary cortices Additional thalamic input to S1cortex arises from the zone incerta both directly viainhibitory projections and indirectly through projec-tions to dorsal thalamus Lastly there are extensivereciprocal intercortical projections among primarysecondary and posteroventral sensory cortices andmotor cortices (Fabri and Burton 1991a Keller1993)

The loops comprising thalamic and cortical fore-brain structures are closed in at least two ways (Fig6)4 Vibrissa S1 cortex sends descending projectionsto the superior colliculus which completes the loopthrough hindbrain and midbrain structures (Fig 3)The dominant descending pathway howeverencompasses the intercortical projection fromvibrissa S1 to M1 cortex Vibrissa M1 cortex sendsdescending projections to the superior colliculusthat as in the case of the descending projectionsfrom S1 cortex complete a loop Importantly thedescending projection from M1 cortex has beenshown to directly overlap with neurons in superiorcolliculus that project to the vibrissa region of thefacial nucleus


FIGURE 5 Forebrain-level sensorimotor loop that projectsdirectly to the midbrain This loop contains a singleforebrain structure the zona incerta (trigeminal gan-glion reg zona incerta) An excitatory projection (Nicoleliset al 1992 Kolmac et al 1998) (zona incerta regtrigeminal ganglion ) An inhibitory connection (Nicoleliset al 1992 Kolmac et al 1998) (zona incerta regsuperior colliculus) An inhibitory connection (Kim etal 1992 Nicolelis et al 1992)

FIGURE 6 Forebrain-level sensorimotor loop The indi-cates the facial motor nerve transiently blocked with anerve cuff in the experiments of Fee et al (1997)(trigeminal nuclei reg thalamus) All trigeminal nucleisend projections to the ventral posteromedial (VPM) andposterior (POm) nuclei in the dorsal thalamus (Lund andWebster 1967 Smith 1973 Erzurumlu and Killackey1980 Mantle-St John and Tracey 1987 Hoogland et al1987 Jacquin et al 1989 Killackey et al 1990 Chiaia etal 1991a Benett-Clarke et al 1992 Diamond et al1992 Nicolelis et al 1992 Williams et al 1994) Therepresentation of the vibrissae forms a somatotopic map(ordf barreloidsordm ) in VPM (Van Der Loos 1976 Sugitani etal 1990) and POm (Nothias et al 1988 Fabri andBurton 1991b) (zona incerta reg thalamus ) Zonaincerta in the ventral thalamus projects to the VPM andPOm nuclei in the dorsal thalamus (Power et al 1999)(thalamus laquo cortex) Thalamic regions VPM POm andzona incerta project to primary (S1) secondary (S2) andposterior ventral areas of sensory cortex and the cortexsends feedback projections to VPM POm and the trigemi-nal nuclei (Wise and Jones 1977 Donoghue et al 1979Donoghue and Kitai 1981 Hoogland et al 1987 Carvelland Simons 1987 Koralek et al 1988 Welker et al 1988Jacquin et al 1990b Chiaia et al 1991a b Diamond etal 1992 Nicolelis et al 1992 Desch Atildeenes et al 1996L Acircevesque et al 1996) The projection from zona incerta tocortex is unique in providing an inhibitory input (Chapin etal 1990 Nicolelis et al 1992) (intercortical) VibrissaS1 cortex forms reciprocal projections with other vibrissasensory areas (Carvell and Simons 1987 Chapin et al1987 Welker et al 1988 Fabri and Burton 1991a) andwith vibrissa motor cortex (White and deAmicis 1977Asanuma and Keller 1991 Fabri and Burton 1991aAroniadou and Keller 1993 Keller 1993 Miyashita et al1994 Izraeli and Porter 1995) The representation of thevibrissae forms a somatotopic map in S1 (Woolsey et al1974 Durham and Woolsey 1977) (ordf barrelsordm ) and S2(Carvell and Simons 1986 Kleinfeld and Delaney 1996)cortices Note that the primary motor cortex is taken as theparasagittal agranular medial area (cortex reg superiorcolliculus) Both sensory and motor cortex send descend-ing projections to the superior colliculus (Wise and Jones1977 Killackey and Erzurumlu 1981 Welker et al 1988Mercier et al 1990) Miyashita and Shigemi (1995) havedemonstrated possible cellular interaction between thedescending cortical M1 projection to colliculus and thecolliculus to facial nucleus projection consistent with therelay of motor commands to the facial nucleus (M1 regsuperior colliculus) A direct connection from thevibrissa motor cortex to an unidentified nucleus in thereticular formation adjacent to the facial nucleus (Miya-shita et al 1994) is suggestive of a central patterngenerator (CPG Fig 2) in analogy with the CPG formastication (Nozaki et al 1986)

Vibrissa M1 cortex sends a direct projection to anundefined nucleus in the reticular formation adja-cent to the facial nucleus (Fig 6) A likely candidateis the parvicellular reticular formation which liesdorsal to the facial nucleus and forms direct projec-tions to the facial nucleus (Mogoseanu et al 1994)This target is an additional candidate for the centralpattern generator that drives rhythmic whiskingmuch as mastication is driven by a pattern generatorin the reticular nuclei (Nozaki et al 1986)

Spike signals in primary cortex during

whisking

All of the published studies to date in awake and eitherbehaving or attentive animals have been at the level ofcortex (Carvell et al 1996 Fee et al 1997) althoughother studies on recordings from both subcortical andcortical areas are in progress (Goldreich et al 1997Moxon et al 1998 Sachdev et al 1998) Part of thebias toward cortex undoubtedly reflects the ease ofaccess to cortical areas across the lissencephalic brainof the rat However justification for this bias is alsoderived from the results of behavioral experimentswith decorticate animals Animals that are devoid ofprimary vibrissa cortex can reorient in response tovibrissa stimulation but fail depth perception tasksthat involve the use of the vibrissae (Hutson andMasterton 1986 Barneoud et al 1991)

EMG as a behavioral measure

The motor output of interest in the study of thevibrissa sensorimotor system is the angle of thevibrissae as a function of time (Fig 1c) While thisposition may be extracted from high-speed videoimages the scalar nature of vibrissa motion duringexploratory whisking (Fig 1a) suggests that a singlemeasure of the activity of the mystacial musculaturemay serve as an accurate correlate of vibrissaposition One such measure is the EMG in whichelectrodes imbedded in the mystacial pad are used torecord the high-frequency spike activity of themusculature (Carvell et al 1991) This signal isrectified and low-pass filtered to yield the extent ofelectrical activation of the mystacial muscles (Kamenand Caldwell 1996) we denote this processed signalthe EMG5 Increases in the amplitude of the EMGcorrespond to protraction of the vibrissae whiledecreases correspond to retraction (Fig 7a)

The EMG measured during exploratory whiskingcontains bouts of rhythmic activity that span 1 ormore seconds as illustrated by the epoch between 2and 4 s in Figure 7a The period of these oscillationsis relatively constant throughout the bout but theenvelope surrounding the oscillations is seen to varyalbeit smoothly over the course of the bout The fastoscillations with a spectral peak near 8 Hz (Fig 7b)correspond to the rhythmic motion of the vibrissaeThe slowly varying envelope with spectral compo-

nents below 5 Hz contains contributions from bothchanges in the maximum amplitude of the whisk andchanges in the midpoint of the vibrissae The lattercorrespondence is illustrated by the data of Carvell etal (1991) who simultaneously measured the mysta-cial EMG and the position of the vibrissae via videoimages and showed that the midpoint of the positiondetermined by the two methods largely track eachother (Fig 7c)

Two forms of rhythmic whisking basic phenomenology

Rats perform at least two forms of rhythmic whiskingthat are correlated with their overall behavioral state(Semba and Komisaruk 1984 Fee et al 1997) Oneform occurs while the animals are immobile andcorresponds to low amplitude tremor-like move-ments of the vibrissae also referred to as ordf twitchingordm(Nicolelis et al 1995) These motions are dis-tinguished by a relatively low amplitude EMG signalwith spectral components between 7 and 11 Hz andhighly synchronous cortical activity as inferred fromthe synchrony of the electrocorticogram (ECoG)(Fig 8a) The second form occurs while animals


FIGURE 7 The electromyogram (EMG) of the mystacialpad and its relation to vibrissa motion (a) A single recordof the EMG that illustrates two bouts of whisking duringan exploratory task (Fig 1a) Note the fast oscillations andthe slowly varying envelope of the oscillations Theamplitude of the EMG is maximum at protractioncorresponding to contraction of the mystacial muscles(Fig 1a) and minimum at retraction which is passive TheEMG was recorded with a 50 m m diameter tungsten wirethat was implanted in the mystacial pad The raw signalwas amplified half-wave rectified and integrated with a4-pole Bessel low-pass filter set at 200 Hz (b) The spectraldensity of the EMG as a function of frequency Note thepeak at 8 Hz which corresponds to the frequency ofrhythmic whisking and the spectral energy at low fre-quencies which corresponds to changes in the envelope ofthis positive definite signal The spectral estimate wasbased on ~ 1000 s of data (c) Relation of the set-point ofvibrissa motion determined from the analysis of videoimages to the envelope of the EMG record The midpointis the angle about which the vibrissae oscillate ((c) isadapted from Carvell et al (1991))

actively explore their environment and correspond tolarge amplitude motion of the vibrissae such as thoseillustrated in Figure 1a These motions are dis-tinguished by a relatively high amplitude EMGsignal with spectral components between 6 and9 Hz and largely asynchronous cortical activity (Fig8b)

Responses in vibrissa sensory cortex during vibrissa

ordf twitchingordm

Chapin Nicolelis and coworkers (Nicolelis et al1995 1997) devised a strategy to simultaneouslyrecord unit activity from multiple regions along thevibrissa sensory branch in rats as well as measure themotion of their vibrissae with video techniques asthe animals rested between periods of exercise Theyobserved highly coherent oscillatory spike activityacross multiple sites in vibrissa S1 cortex as well assites in the principal and spinal trigeminal nuclei(PrV and SpV respectively in Figure 9a) and indorsal thalamus (VPM in Figure 9a) The phase ofthe oscillations was relatively uniform both withinand between brain regions such that most of thevariance of these responses was accounted for by thefirst principal component of the temporal response(Fig 9a) Further the responses were time-locked tolow amplitude tremor movements of the vibrissae(Fig 9b) Thus all units in primary vibrissa cortexhave a similar preferred phase within the whisk cyclefor spiking

The nature of the cortical activity and the largecoherence of this activity with the vibrissa movementare reminiscent of the ordf tremorordm state described bySemba and Komisaruk (1984) (Fig 8a) The brainactivity in this state is thought to be dominated bythalamocortical oscillations variously described asordf high-voltage spindlesordm (Buzsaki 1991) or absentseizures (Steriade et al 1993 McCormick and Bal1997) The data of Nicolelis et al (1995) areremarkable in that they show that such coherentactivity extends to subthalamic areas However as

the vibrissa sensorimotor pathway forms a closedloop (Fig 6) it is impossible to determine if theoscillatory activity in the trigeminal nuclei resultsfrom feedback connections from higher areas or fromperipheral reafference


FIGURE 8 Vibrissa and cortical activity during two typesof whisking (a) The tremor or vibrissa ordf twitchingordm stateNote the relatively low amplitude of the right (R) and left(L) electromyograms (EMGs) and the high degree ofcortical synchrony in the frontal (F) and occipital (O)electrocorticograms (ECoG) (b) The exploratory whisk-ing state Note the relatively high amplitude EMG and theloss in cortical synchrony (adapted from Semba andKomisaruk (1984))

FIGURE 9 Simultaneously recorded brain activity frommultiple sites in the brainstem trigeminal nuclei (PrVprincipal trigeminal nucleus SpV spinal trigeminalnuclei) thalamus (VPM ventral posterior medial nucleusof dorsal thalamus) and vibrissa primary sensory (S1)cortex during epochs of vibrissa ordf twitchesordm (a) Single-trialspike rasters plots from separate extracellular unitsgrouped according to anatomical area for a particular 5 sinterval The lower smooth trace is the first principalcomponent of the raster responses defined as the leadingeigenvalue of the covariance matrix that is formed bysumming the outer products of the individual responses(Ahmed and Rao 1975) (b) The relation of vibrissaposition determined from video images to the time- andsite-averaged spike responses Note that the spike activityin each region has on average a single preferred phase inthe whisk cycle (adapted from Nicolelis et al (1995))

Single-unit responses in vibrissa sensory cortex during

exploratory whisking

We now turn to an understanding of the spike trainsin vibrissa S1 cortex during an exploratory whiskingtask Fee et al (1997) devised a paradigm in whichrats were trained to perch on a ledge and whisk insearch of a food tube The rats performed this taskwith blindfolds on during which time they whiskedover large angles for epochs of ~ 5 s (Fig 1a) Thepositions of the vibrissae were inferred from meas-urements of the EMG Extracellular signals weresimultaneously recorded from multiple sites in S1cortex during the whisking epochs from whichsingle-unit spike trains (Fee et al 1996) as well aslocal field potentials (LFP) were derived No sig-nificant correlation between the spike trains recordedfrom different locations was found and as discussedin detail later (Fig 13) the LFP was featurelessduring the epochs of whisking The large rhythmicwhisking motions and the asynchronous extracellularactivity imply that the animal is in the ordf exploratorywhiskingordm state of Semba and Komisaruk (1984)(Fig 8b)

Fee et al (1997) found a significant correlationbetween the occurrence of a spike and the phase ofthe vibrissae within the whisk cycle (Fig 1c) asillustrated by the three examples in Figure 10aImportantly different units tended to spike atdifferent phases so that each unit could be asso-ciated with a preferred phase This is in contrast tothe case of vibrissa ordf twitchingordm in which all unitstended to fire at the same phase (near full protrac-tion Fig 9b) The distribution of the modulationdepth and the phase of the correlation over allmeasured units shows that the values of preferredphase are uniformly represented but that there is atendency for the modulation depth to be greatest forunits whose preferred phase corresponds to protrac-tion from the retracted position (Fig 10b)

The above results show that spike activity iscorrelated with relative position within the whiskcycle Fee et al (1997) also found significantcorrelations between the envelope of the rhythmicwhisking epochs (Fig 1c) and the single-unit spiketrains The envelope was found by demodulating theEMG near the whisking frequency (insert Fig 11)so that the values of the envelope correspond to theamplitude of the rhythmic whisking motion Thedistribution of amplitudes at all time points wascompared with the distribution of amplitudes attimes when a spike occurred For the example ofFigure 11 we observe a significant decrement in theoccurrence of a spike at small amplitudes of theEMG with a concomitant enhancement at largeamplitudes Overall relative enhancement or sup-pression of the spike rate at large whisking amplitudeoccurred with essentially equal probability Therewas no obvious relationship between the degree ofamplitude coding (Fig 11) and phase modulation

(Fig 10) among the single units in theseexperiments

How well can vibrissa position as inferred fromthe EMG signal be predicted from the spike trainof a single unit The linear transfer function thatpredicts the EMG signal from the spike train was


FIGURE 10 Relation of single-unit spike activity in vibrissaS1 cortex to vibrissa position (a) Correlations betweenthree simultaneously recorded units and the electromyo-gram (EMG) The correlations were calculated as spikeaverages triggered off the positive peak of the EMG Notethat the phase is different for each unit The modulationdepth of the correlation is defined as the peak-to-peakamplitude divided by the baseline spike rate equal to thatat long lag times The phase corresponds to 2 p -times thetime-lag to the peak in the correlation divided by theperiod of the oscillation (b) Polar plot of the modulationdepth and preferred phase for 115 units across threeanimals different symbols refer to different rats Note thatall preferred phases are represented (adapted from Fee etal (1997))

calculated as an average over all but one trial for aparticular data set (Fee et al 1997) The transferfunction shows damped oscillations near the whiskfrequency (Fig 12a) When applied to the spiketrain of the excluded trial the predicted EMGsignal appears to coincide well with the measuredsignal (Fig 12b) The ability to predict the EMGfrom spike trains implies that the output of someunits has both a sufficiently high probability ofspiking and a sufficiently strong correlation with theEMG so as to represent vibrissa position on a real-time basis

We now consider the direction of signal flow in theforebrain sensorimotor loop (Fig 6) The correlatesof vibrissa position in S1 cortex may originate eitherfrom peripheral reafference or discharge along acentral pathway eg feedback from vibrissa M1cortex To determine the pathway(s) Fee et al

(1997) used a transient block to the facial nerve toordf openordm the sensorimotor loop ( in Fig 6) Theanimals continued to whisk on their ipsilateral sideduring the block the normally high coherencebetween whisking on both sides of the face allowedthe ipsilateral EMG to serve as the measuredreference of vibrissa position The conclusion fromthese manipulations was that the fast correlate ofvibrissa position (Fig 1c) which reports the phase ofthe whisk cycle (Fig 10) and varies on the 1plusmn 10 mstimescale originates via peripheral reafference Incontrast the slow correlate of vibrissa position

(Fig 1c) which reports the amplitude of the whisk(Fig 11) and varies on the 01plusmn 1 s timescaleoriginates from a corollary discharge

Two forms of rhythmic whisking correspondence with

synchronous vs asynchronous activity in S1 cortex

In the course of the exploratory whisking experi-ments (Figs 10 plusmn 12) there were epochs of time inwhich the animals rested before resuming the taskDuring the rest period the behavior of the ratalternated between immobility and exploration Wefocus on the record of Figure 13 In the early part ofthe record the state of the animal included epochs ofstrong LFP activity near 10 Hz which were alwaysaccompanied by vibrissa ordf twitchesordm and weak EMGactivity as well as epochs of asynchronous LFPactivity some of which were accompanied by epochsof a strong EMG (Spindling or Whisking Fig 13) Aprotracted period of synchronous LFP activity coin-cided with the appearance of weak vibrissa ordf tremorsordm(Spindling Fig 13) Lastly there was a completeabsence of synchronous LFP activity but the appear-ance of a strong EMG signal at the whisk frequencyas the animal performed the task (ExploratoryWhisking Task Fig 13) These data imply that thetwo forms of whisking (Semba and Komisaruk1984) ordf twitchesordm (Figs 8a and 9) vs exploration(Figs 1 8b and 10 plusmn 12) coincide with mutuallyexclusive states of cortical activation


FIGURE 11 Relation between the amplitude of the whiskcycle and probability of spiking for a particular single unitin vibrissa S1 cortex The amplitude of the electromyogram(EMG) envelope denoted V is sampled at each time point(insert the solid line indicates the envelope and the dotsdenote values of the envelope that coincide with a spike)We plot the number of occurrences of the value of theEMG envelope at all times (Values at All Sample Times)and at times when a spike occurred (Values at SpikeTimes) The difference between the two curves resultsfrom a correlation between EMG amplitudes and theprobability of spiking The significance of the difference isestablished by a Kolmogorov plusmn Smirnov test (adapted fromFee et al (1997))

FIGURE 12 The transfer function or linear predictor ofthe electromyogram (EMG) from the measured spiketrain (a) Plot of the linear predictor calculated as a least-squares estimator over 19 of 20 trials of simultaneouslyrecorded spike trains and EMGs Note that the transferfunctions decay in about one whisk period (b) Compar-ison between the measured EMG and the predicted EMGcalculated as a convolution of the transfer function in part(a) with the measured spike train The predicted EMGsignal is seen to coincide well with the measured signal aregion of poor prediction at the beginning of the trialcoincides with the unit not firing ((b) is adapted from Feeet al (1997))

Single-unit responses in vibrissa motor cortex during

whisking

Vibrissa primary motor cortex forms extensivereciprocal projections with S1 cortex (Fig 6) A

priori these connections could lead to shared infor-mation about actual and intended vibrissa positionTo measure the cortical correlates of vibrissa positionin M1 Carvell and coworkers (1996) devised aparadigm in which gentled but otherwise naive ratswere enticed to whisk in response to an attractiveodorant The whisks encompassed large angles andwere sustained for epochs of ~ 1plusmn 5 s The position ofthe vibrissae were inferred from measurements of theEMG and extracellular signals were recorded fromsites in M1 cortex during the whisking epochs (Fig14a) Carvell et al (1996) observed that unlike thecase of vibrissa S1 cortex (Fig 10) there were nostrong correlations of unit output with the rhythmicwhisking pattern6 In contrast as in the case ofvibrissa S1 cortex (Fig 11) Carvell et al (1996)found a strong correlation between unit activity andthe envelope of EMG activity (Fig 14b) In thesemeasurements the envelope contained contributionsfrom both a change in set-point of the vibrissae anda change in the amplitude of the whisk

Transfer functions for rhythmic signaling

along the motor branch

The results of Carvell et al (1996) imply that the fastcortical reference signal of vibrissa position (Fig 10)is not transmitted between S1 and M1 cortices

despite the direct connections between these areasand the ability of even layer 1 intercortical connec-tions to drive spiking in neurons throughout thetarget column (Nakajima et al 1988 Cauller andConnors 1994) One question suggested by thisresult concerns the nature of neuronal control signalsalong the motor branch of the sensorimotor pathway(Fig 6) ie M1 reg superior colliculus reg facialnucleus reg vibrissae We thus consider preliminarydata (R W Berg and D Kleinfeld unpublished) onthe temporal response properties along this branchThe approach in these experiments was to quantifythe movement of the vibrissae that was elicited byrepetitive microstimulation of different brain regionsin the anesthetized animal

Microstimulation was used to activate a brainregion with a train of pulsed constant-current stim-uli The trains were constructed so that a singlestimulus could elicit vibrissa motion (see legend toFigure 15) as described in mapping studies based onmicrostimulation (Hall and Lindholm 1974Gioanni and Lamarche 1985 Neafsey et al 1986Donoghue and Sanes 1988 Sanes et al 1990Miyashita et al 1994 Weiss and Keller 1994) Themotion of the vibrissae was monitored magnetically(Fig 15a) We calculated the value of the transferfunction between the repetitive stimuli and themeasured motion of the vibrissae at the stimulationfrequency

We observed that at the level of the facial nucleusthe vibrissae could accurately follow stimulation upto a roll-off of ~ 12 Hz (Fig 15b) above the ~ 8 Hzfrequency of exploratory whisking (Fig 7b) In


FIGURE 13 Relation of cortical synchrony to the electromyogram (EMG) during immobile vs exploratory states in the ratThe top spectrogram is the power density in the EMG vs frequency and time A strong peak in the spectrogram near 10 Hzcoincides with exploratory whisking while a weak or undetectable peak coincides with immobility including epochs ofvibrissa ordf twitchesordm Note that increased power density is coded by lighter shades of gray During the epoch labeledordf Spindlingordm the vibrissae were observed to make fine tremor or ordf twitchordm movements (video data not shown) Thebehavioral markers during the interval marked ordf Exploratory Whisking Tasksordm indicate when the animal successfullycompleted the search task with its vibrissae The bottom spectrogram is the power density in the local field potential (LFP)Note that a peak in this spectrogram coincides with synchronous electrical activity in S1 cortex Such a peak occurs duringspindling but not during exploratory whisking consistent with the data in Figure 7

contrast at the level of M1 cortex the vibrissaecould follow stimulation only up to a roll-off of ~3 Hz (Fig 15c) well below the frequency of explora-tory whisking An intermediate frequency responsewas observed at the level of the superior colliculus(data not shown) These results imply that M1 cortexcannot control vibrissa position of the fast timescaleof the whisk cycle but can control motion on longertimescales One caveat is that the timescales may bedifferent in awake attentive animals particularly atthe level of M1 cortex

Timescales of signaling in the sensorimotor

loop

The electrophysiological data on the sensory andmotor branches of the vibrissa sensorimotor loop(Figs 10 11 14 and 15) can be combined with theanatomical structure of the forebrain loop (Fig 6) toyield an overview of signal flow (Fig 16) We observe

fast signaling that starts at the facial nucleus andcourses through the vibrissae and up through thetrigeminal pathway to thalamus and S1 cortex Theaddition of a slow signal that correlates with the


FIGURE 14 Relation of unit spike activity in vibrissa M1cortex to vibrissa position (a) Simultaneous records of asingle-trial spike train and the concomitantly recordedelectromyogram (EMG) Here as in all cases there is noobvious correlation between the occurrence of a spike andthe amplitude of the EMG (b) Plots of the trial-averagedspike response for a particular unit and the average EMGThe averages were aligned to start 10 s before the onset ofwhisking Note the clear correlation between the rate ofspiking and the value of the EMG (adapted from Carvell etal (1996))

FIGURE 15 Preliminary data on the ability of the vibrissaeto follow rhythmic stimulation of different areas in themotor branch of the sensorimotor loop Long Evans femalerats were anesthetized with ketaminexylazine (0005(ww) and 00015 (ww) respectively) and held ster-eotaxically Constant-current stimuli (suprathreshold uni-polar 200 m s wide pulses of 10 plusmn 50 m A that were repeatedfive times at 2 ms intervals) were delivered throughtungsten microelectrodes (1plusmn 2 M V at 1 kHz) Vibrissaposition was determined electronically by detecting themotion of a small magnet glued to the principal vibrissawith a magnetoresistive bridge The magnet plus extraglue to stiffen the shaft increased the moment of inertia ofthe vibrissa by ~ 50 We calculated the transfer functionbetween a train of stimuli and the vibrissa motion for arange of stimulation frequencies the transfer function wastypically statistically significant at the stimulation fre-quency and at its harmonics (a) Trace of the position of avibrissa as the facial motor nucleus was stimulated at arepetition rate of 83 Hz (b) Plot of the magnitude of thetransfer function at the value of the stimulation frequencyas a function of stimulation frequency for the ipsilateralfacial motor nucleus Note the onset of roll-off near 15 Hz(arrow head) with slope f plusmn 2 The bars are one standarddeviation calculated as a trial-to-trial average (c) Bodeplot of the magnitude of the transfer function at thestimulation frequency for the contralateral vibrissa M1cortex Note the onset of roll-off near 3 Hz (arrow head)with slope f plusmn 4

envelope of rhythmic whisking is manifest by S1Slow signaling is manifest along the motor branchdown to the level of the facial nucleus to completethe loop At present it is unknown if the slow corticalreference signals for vibrissa position that areobserved in S1 and M1 cortices are in fact relatedalthough the central origin of the slow signal in S1(Fee et al 1997) suggests that that may be soFurther it is unknown if and how the fast and slowsignals interact in the sensation of objects by thevibrissa as may be posited to occur in S1

Detection of vibrissa contact

We shift our focus and consider how the rat maydetermine the angle of the vibrissa relative to theorientation of the mystacial pad upon their contactwith an object A central and yet unresolved issue iswhether the output of neurons in S1 is ordf tunedordm to theposition of the vibrissa during contact In otherwords does the probability that a unit may spikeupon contact vary as a function of the angle of thevibrissae in the whisk cycle To the extent that

activation of the sensory afferent depends on achange in the position of a vibrissa relative to that ofits follicle (Fig 1b) we assume for purposes ofdiscussion that activation is equally likely at all anglesin the whisk cycle The issue then becomes one ofdeducing the angle of contact under two assump-tions (i) a neuron may fire at a preferred phase of thewhisk cycle in the absence of contact to form areference signal (Fig 17a see also Figure 10b) and(ii) a neuron may spike upon contact with an objectduring any phase of the whisk cycle (Fig 17b)

There are two general classes of decoding schemesfor reconstructing the phase of the vibrissae uponcontact with an object In the first class the output ismaximal when the equal-time correlation between thetwo sensory signals is maximal ie when the phase ofcontact is close to that of the reference signal (Fig17c) This scheme may be implemented in a variety offunctionally equivalent ways (Kam et al 1975) Inone realization the inputs from the two sensorysignals are summed and the sum causes a neuronaldecoder to fire only if it exceeds a judiciously setthreshold value In another the sensory inputs aremultiplied (Koch and Poggio 1992) and the productmodulates the output of a neuronal decoder In thesecond class of decoding schemes the output isminimal when the phase-shift between the two sensorysignals is zero ie when the phase of contact is closeto that of the reference signal (Fig 17d) This scheme


FIGURE 16 Summary schematic of the timescale ofsignaling along the sensorimotor loop that involves thecortex (Fig 6) The dark arrows correspond to pathwaysthat were directly or indirectly probed as part of the studiesreviewed here Specifically fast signaling of vibrissa posi-tion through the sensory branch is known from directmeasurements with anesthetized and sessile animals (tri-geminal ganglion (Simons et al 1990) trigeminal nuclei(Nicolelis et al 1995) thalamic nuclei (Carvell andSimons 1987 Diamond et al 1992 Nicolelis et al 19931995) S1 (Welker 1976 Simons 1978 Armstrong-Jamesand Fox 1987 Armstrong-James et al 1992 Nicolelis etal 1995) and by inference from the presence of a fastcortical correlate of vibrissa position in S1 (Fee et al1997) (Figs 8 plusmn 10) Slow signals of vibrissa amplitudeswere established in S1 and M1 from measurements withawake animals (Carvell et al 1996 Fee et al 1997) (Figs11 and 14) Slow control signaling along the motor branchincluding M1 and the superior colliculus is implied by theresults of microstimulation experiments (Fig 15) Sim-ilarly fast control of vibrissa position by the facial nucleuswas established by microstimulation experiments (Fig15) The light gray arrows refer to major paths in thesensorimotor loop with uncharacterized signalingproperties

FIGURE 17 Diagram of decoding schemes to extract therelative phase of the vibrissae upon their contact with anobject (a) The probability of a unit in S1 producing a spikefor whisking in air This tuning curve defines the preferredangle for the fast reference signal (Fig 10) (b) Theprobability of a unit in S1 producing a spike upon contactof the vibrissa with an object It is assumed that the contactoccurs at an angle C relative to the preferred angle Thefree whisking frequency is n D (c and d) The spiking outputas a function of contact angle relative to the preferred anglefor different decoding schemes The scheme in (c) is basedon the correlation of the two inputs and produces amaximal output when contact occurs at the preferred angleof the reference signal The scheme in (d) is based on aphase comparison of the two inputs and produces aminimal output at the preferred angle and a maximaloutput away from the preferred angle of the referencesignal

occurs naturally in phase-sensitive detection and isdiscussed in detail below We emphasize the dramaticand experimentally testable difference in the predic-tions between these two classes of decoding schemeone yields a maximal signal at the preferred referencephase of neurons in S1 while the other yields aminimal signal (cf Figs 17c d)

Phase-sensitive detection of vibrissa contact

Ahissar et al (1997) addressed the question of howvibrissa contact is coded by neurons in S1 cortexThey presented evidence that this sensory processmay involve a phase-sensitive servo mechanism toextract the position of the vibrissae upon theircontact with an object (Fig 17d)

We consider first a brief tutorial on the use ofphase-sensitive detection to lock the frequency of aninterval oscillator with that of an external rhythmicinput (de Bellescize 1932) A phase-sensitive servoconsists of two parts that are connected by afeedback loop (Fig 17a) (i) an input controlledoscillator which may be an intrinsic neuronal oscil-lator whose frequency is monotonically shifted by anexternal rhythmic input and (ii) a comparator thatgenerates an output signal that is proportional to thedifference in timing of the intrinsic neuronal oscil-lator and that of an oscillatory external signal Theoutput from the comparator may need to be filtereddepending on the details of the servo and is used asan error signal to shift the frequency of the intrinsicneuronal oscillator so that it matches the frequencyof the external reference signal7 The error signal is asensitive indicator of changes in the timing both infrequency and phase of the input

In steady state the frequency of the intrinsicneuronal oscillator denoted n R is locked to thefrequency of a rhythmic external drive denoted n D that is applied to the vibrissae ie

n R = n D (Eqn 1)

Even though the frequencies are locked there maybe a constant difference in the timing between theintrinsic neuronal oscillator and the external driveThis is expressed in terms of the phase difference f between the output of the two oscillators ie

f = f plusmn 1 5 n D plusmn n Rfree

G 6 + 2 p n D( t D plusmn t R ) (Eqn 2)

The first term in eqn 2 is a function of the errorsignal from the comparator denoted f(x) whichdepends on the frequency of the external rhythmicinput relative to the natural frequency of the intrinsicneuronal oscillator n R

free and on the gain of theintrinsic neuronal oscillator G (Fig 18a) Thesecond term is a function solely of signal delayswhere t D is the peripheral delay taken as t D = 10 ms


FIGURE 18 Model and predictions based on the modelfor a neuronal phase-sensitive servo mechanism (a)Schematic of a servo that includes major delays insignaling A comparator and filter generates a slowlyvarying output (error signal) that is a function of thedifference in timing between input from the corticalreference of rhythmic whisking (with drive frequency n D )and the intrinsic cortical oscillator (with frequency n R =n R

free + G (error signal)) For consistency with theexperimental data (Fig 19c) we consider a comparatorwhose error signal is a periodic function of the inputtiming difference with a frequency of 2 n D Note that theobserved delay between the intrinsic cortical oscillationsand the rhythmic drive to the vibrissae t OS includescontributions from peripheral delays t D and centraldelays t R intrinsic to the servo (b) Diagram thatillustrates how the phase of the intrinsic neuronal oscil-lator is transiently modulated upon contact with anobject during rhythmic whisking Prior to contact thereference of rhythmic whisking occurs at a fixed phaseUpon contact of the vibrissae with an object at an angleof just over p 2 in the diagram the phase of the externalrhythmic reference is shifted as spikes occur at both thecortical reference position and on contact with an objectThe change in average angle leads to a transient shift inthe phase of the intrinsic cortical oscillator that dependson the phase of contact relative to the preferred phasefor the cortical reference ie C in the diagram (see alsoFigure 17d) An alternative servo scheme is given byAhissar et al (1997)

(Armstrong-James et al 1992) and t R is a centraldelay

One function of a phase-sensitive servo is togenerate a transient change in the error signal fromthe comparator when the phase of the input ischanged This will occur when the vibrissae contactan object during rhythmic whisking as illustrated inFigure 18b and thus the phase-sensitive servoprovides a mechanism to explain the ordf phasicordm on-and off-responses of neurons Prior to contact thephase of the external rhythmic input is constant as inFigure 9a Upon contact of the vibrissae with anobject the phase of the rhythmic whisking reference

is shifted as spikes occur at both a reference positionand at object contact within the whisk cycle Thisleads to a transient shift in the phase of the intrinsicneuronal oscillator (Fig 18b) The duration of thistransient is expected to be at least as great as theinverse of the oscillation frequency ie 1 n D orhundreds of milliseconds The amplitude of thetransient is minimal if contact occurs at the preferredphase of the reference signal and increases as contactoccurs before or after the preferred phase (Figs 17dand 18b)

Cortical units that behave as intrinsic neuronaloscillators in the absence of sensory input were


FIGURE 19 Data and analysis in support of a model forphase-sensitive detection of rhythmic drive of the vibrissae(a) Intrinsic neuronal oscillators in vibrissa S1 cortex thatshift their frequency in the presence of iontophoreticallyapplied glutamate Shown are the autocorrelation func-tions for spontaneous activity (panels 1 and 3) and activityin the same unit in response to glutamate (panels 2 and 4)glutamate is seen to raise the frequency of the oscillatoryfiring The summary plot shows that in general the firingrate is monotonically increased in the presence of thistransmitter (b) Evidence that the intrinsic neuronaloscillators can frequency lock to rhythmic stimulation ofthe vibrissae Shown are the interspike interval distribu-tions during spontaneous activity (panels 1 3 and 5) andthe change in these distributions in response to externalvibrissa stimulation at 5 Hz (panel 2) and 8 Hz (panel 4)Note that the frequency remains partially entrained afterremoval of the stimulus (cf panels 2 and 3 and 4 and 5)(c) The time delay t OS between the onset of externalstimulation and that of spiking in the intrinsic oscillatorsThe data were pooled into two populations based on thenature of the response Note the apparent reset in the delayfor the major population near a stimulus frequency of13 Hz (d and e) The measured time delays in (c) arereplotted as phase vs frequency Note the offset of twopoints in (d) by the addition of a phase of + p radians Thesolid line is a fit to eqn 2 with f (x) = x and parametersindicated in the figure The periodicity of p radians in thedata of (d) suggests that the output of the comparator (Fig18a) has a periodicity of 12vD as opposed to 1vD for datawith a periodicity of 2 p radians ((a) plusmn (c) are adapted fromAhissar et al (1997))

demonstrated in anesthetized animals by Dinse et al

(1992) Ahissar et al (1997) recorded the extrac-ellular spike signal from similar units and showedthat the frequency of the oscillations which rangedbetween 5 and 10 Hz in the absence of externalinput is monotonically increased by the local appli-cation of glutamate an excitatory neurotransmitterThe shift in frequency was reversible and observedacross all cells tested (Fig 19a) These cells satisfythe role of intrinsic neuronal oscillators whosefrequency may be monotonically shifted

Neurons that serve as reporters of a rhythmicallyvarying external reference signal were described byFee et al (1997) (Fig 10a)8 Neurons that comparethe timing of signals from an intrinsic neuronaloscillator in cortex (Fig 18a) with that of a referenceof vibrissa motion remain unidentified (Simons andHartings 1998) although work in progress (Ahissarand Haidarliu 1998) suggests that such neurons maylie in the posterior nucleus of dorsal thalamusconsistent with the slow signaling in the paral-emniscal path (Diamond et al 1992) NonethelessAhissar et al (1997) found that the output of theintrinsic neuronal oscillators they identified (Fig19a) would frequency lock to a rhythmic driveapplied to the vibrissae (Fig 19b) The time delaybetween the intrinsic neuronal oscillators and theexternal drive to the vibrissae t O S (Fig 18a) wasnot constant as expected for units that are solelydriven by an outside stimulus (circles Fig 19c) butvaried systematically with the frequency of therhythmic drive (squares Fig 19c) This variation is asignature of a phase-sensitive servo In fact thesedata are well fit by a straight line when replotted asphase ( f = 2 p n D t O S Fig 19d) The slope andintercept of this line are accounted for by reasonableparameters in the model (eqn 2 with for simplicityf(x) = x Fig 19d) similar parameters accounted forthe units that were solely driven by outside stimula-tion (Fig 19e)

The good fit between the data and a straight linepredicted by the model (eqn 2) is enticing but servesonly as a consistency argument based on data largelyfrom anesthetized animals The validity of the phase-sensitive model as the mechanism used by theforebrain to compute vibrissa position relative to themystacial pad awaits further experimental evidenceparticularly with regard to the identification ofcandidates for the comparator circuit and the natureof the output from these circuits (Fig 17d)

Concluding remarks

Extensive anatomical studies have shown that thecircuitry involved in the vibrissa sensorimotor systemis configured as a series of closed loops (Figs 2 plusmn 616) These anatomical studies suggest the need forelectrophysiological measurements that delineate thefunctional roles of the pathways At the level of thehindbrain (Fig 2) the trigeminal nuclei provide a

link between sensory input and motor control It willbe instructive to determine if there is a sensory-mediated control of motor output at the level of theordf vibrissa reg trigeminal ganglia reg trigeminal nucleus

reg facial nucleus reg vibrissaordm loop possibly throughsimultaneous recordings from the trigeminal andfacial nuclei during a behavior task A second issue isto identify the putative central pattern generator ofrhythmic whisking

At the level of the midbrain (Fig 3) the superiorcolliculus receives convergent inputs via trigeminalthalamic cortical basal ganglion and cerebellarpathways (Figs 4 plusmn 6) and is thus positioned tointegrate direct and indirect sensory channels withmotor commands Despite this rich convergence ofpathways there is a dearth of information onsignaling in this structure in the awake rat Thepresence of closed-loop connections between thesuperior colliculus and cerebellar nuclei (Fig 4)raises the further possibility that feedback betweenthe superior colliculus and the cerebellum couldform a central pattern generator for rhythmic whisk-ing simultaneous measurements of olivary oscilla-tions and whisking may settle this issue Lastly therole of the pontine-cerebellar pathway (Fig 4) inmediating the behavior of animals in response tovibrissa contact is unexplored although experimentson the lick response in rats (Welsh et al 1995)suggest the strong involvement of cerebellar nuclei

Multiple thalamic and cortical structures areinvolved in forebrain pathways that merge vibrissasensory input with motor control (Figs 5 and 6) andthe results of behavioral experiments imply that thispathway must be intact for animals to pair a sensoryinput with a learned percept (Hutson and Mas-terton 1986) The issue of how animals compute theabsolute position of their vibrissae as a step towardconstructing a stable percept of objects in theirenvironment has been only partially addressed Onthe one hand the spike trains of some units invibrissa S1 cortex accurately code vibrissa position ina single trial (Fig 12) On the other hand there isincomplete evidence as to how the rat uses thisinformation In the scheme of Ahissar et al (1997)the rat computes the angle of vibrissa contact basedon the relative timing between the reference signaland that caused by contact (Figs 17d 18 and 19)Alternatively the sensitivity of cortical units couldvary as a function of their angle in the whisk cycle inanalogy with position preference in the visual system(Andersen and Mountcastle 1983) (Fig 17d) Itshould be possible to distinguish among thesepossibilities by recording from animals as theirvibrissae are perturbed at different angles in thewhisk cycle during an exploratory whisking task

The experimental evidence shows that the repre-sentation of vibrissa position in cortex involves bothfast signals which can vary on the 10 ms timescaleand report the phase of the vibrissae within a whiskcycle (Fig 10) and slow signals which vary on the


1 s timescale and report the amplitude andor set-point of the whisk (Figs 11 and 14) The fast signaloriginates from peripheral reafference but it isunknown how such reafference results in differentunits spiking at different phases in the whisk cycleThe slow signal originates centrally but the source ofthis signal and the possible relation of this signal tothe fast reference is unknown It is tempting tospeculate that a phase-sensitive servo is a linkbetween the fast and slow cortical reference signalsso that the output of a comparator ie an errorsignal (Fig 18a) is the slow cortical reference signalthat correlates with the set-point of the vibrissa (Figs7 and 14) This correspondence implies that thefrequency of exploratory whisking will change as theanimal changes its set-point eg as it protracts itsvibrissae to palpate an object the behavioral data ofCarvell and Simons (1990) are consistent with thisidea

Acknowledgements

This article is based on material presented atordf Barrels XIordm We thank Ford Ebner for suggestingthe topic of the presentation We further thankMichale Fee for contributing to the data in Figures 7and 13 Suri Venkatachalam for contributing to thedata in Figure 15 Ehud Ahissar James Bower JohnChapin Asaf Keller and Daniel Simons for perti-nent discussions Beth Friedman and two anony-mous referees for their comments on an earlierversion of the manuscript and the Whitehall Foun-dation for their support

Notes

1 Anatomical abbreviations Hindbrain TG trigeminalganglion TN trigeminal nuclei including PrV princi-palis SpVO oralis SpVI interpolaris SpVC caudalisFN facial nucleus Midbrain SC superior colliculusThalamic nuclei VPM ventral posteromedial POmposterior ZI zona incerta VL ventral lateral CortexS1 primary sensory S2 secondary sensory PVposteroventral M1 primary motor cortex

2 The central pattern generator hypothesis derived pri-marily from experimental evidence gathered with inver-tebrates and subcortical vertebrate preparations statesthat rhythmic motions are controlled by autonomousneuronal circuits albeit circuits whose output may begated or modulated by extrinsic inputs (Kleinfeld andSompolinsky 1988 Getting 1989)

3 For completeness we note that the basal ganglia alsoform a closed loop with the superior colliculus Thecolliculus sends projections to the basal ganglia(Tokuno et al 1994) and the basal ganglia sendinhibitory connections to the colliculus (Faull andMehler 1978 Gerfen et al 1982 Westby et al 19931994 Niemi-Junkola and Westby 1998) the role of thebasal ganglia in whisking is presently unclear

4 We note two additional loops at the level of theforebrain One involves projections between the fore-brain and the cerebellum Vibrissa S1 cortex sendsdescending projections to the pons (Kennedy et al1966 Wise and Jones 1977 Shambes et al 1978

Wiesendanger and Wiesendanger 1982a b Welker etal 1988 Mercier et al 1990) whose climbing fiberinput to the cerebellum overlaps with that from thetrigeminal nuclei (Bower et al 1981 Morisette andBower 1996) Vibrissal M1 cortex also projects to thepons (Wiesendanger and Wiesendanger 1982a bMiyashita et al 1994) Feedback to cortex occurs viaprojections from the deep cerebellar to ventral lateralthalamic nuclei (Haroian et al 1981 Gerfen et al1982 Deniau et al 1992 Middleton and Strick1997) The second loop involves projections fromvibrissa S1 and M1 cortices to basal ganglia (Wise andJones 1977 Donoghue and Kitai 1981 Mercier et al1990 L Acircevesque et al 1996) Feedback to cortex occursvia ventral lateral thalamus as in the case of cerebellarfeedback (Fig 6)

5 The definition of the EMG is ambiguous in theliterature Some authors refer to the unprocessed signalof muscular spikes as the EMG and to the rectified andfiltered signal as the rectified EMG while others referto the rectified and filtered signal as the EMG We takethe latter definition and note that as a practical matterone typically records only the relatively slowly varyingrectified and filtered signal

6 In the experiments of Carvell et al (1996) theanimals were untrained so that only a limited numberof trials could be obtained It may be useful to repeatthese measurements with trained animals as well asuse phase-sensitive methods to look for weakcorrelations

7 The output of the comparator is a periodic function ofthe difference between the time dependence of the twoinput signals ie it is of the form f2 p n R t plusmn 2 p n R t D +f plusmn (2 p n D t plusmn 2 p n D t D ) which equals f2 p n D( t R plusmn t D )+ f at lock While the periodicity of this function isover an interval of 2 p in typical realizations of a phase-sensitive servo only a periodicity of p is consistent withthe data of Ahissar et al (1997) This suggests that theputative neuronal comparator doubles the frequency ofthe reference from n D to 2 n D A neural mechanism thataccomplishes this is a summation of the on- and off-responses from the same oscillatory input

8 Since different units in vibrissa S1 cortex have differentpreferred phases (Fig 10) there are presumably multi-ple copies of the comparator circuit An alternativepossibility is that there is a single comparator circuitthat uses a single reference signal that is determined asthe vector average of the single unit responses For thedata of Figure 10b the vector average points towardthe retracted but protracting position


References

AHISSAR E and S HAIDARLIU (1998) Thalamic oscilla-tions compute the phase difference between vibrissalinput and barrel cortex oscillations Soc Neurosci Abstr24 151

AHISSAR E S HAIDARLIU and M ZACKENHOUSE (1997)Decoding temporally encoded sensory input by corticaloscillators and thalamic phase comparators Proc NatlAcad Sci USA 94 11633plusmn 11638

AHMED N and KR RAO (1975) Orthogonal Transforms forDigital Signal Processing Springer New York

ANDERSEN RA and VB MOUNTCASTLE (1983) Theinfluence of angle of gaze upon the excitability of thelight sensitive neurons of the posterior parietal cortex JNeurosci 3 532 plusmn 548

ARMSTRONG-JAMES M and K FOX (1987) Spatio-temporal convergence and divergence in the rat S1ordf barrelordm cortex J Comp Neurol 263 265 plusmn 281

ARMSTRONG-JAMES M K FOX and A DAS-GUPTA

(1992) Flow of excitability within barrel cortex onstriking a single vibrissa J Neurophysiol 681345plusmn 1358

ARONIADOU VA and A KELLER (1993) The patterns andsynaptic properties of horizontal intracortical connec-tions in the rat motor cortex J Neurophysiol 701493plusmn 1553

ARVIDSSON J (1982) Somatotopic organization of vibris-sae afferents in the trigeminal sensory nuclei of the ratstudied by transganglionic transport of HRP J CompNeurol 211 84plusmn 92

ASANUMA H and A KELLER (1991) Neural mechanismsof motor learning in mammals NeuroReport 2217plusmn 224

BARNEOUD P M GYGER F ANDRES and H VAN DER

LOOS (1991) Vibrissa-related behavior in mice transienteffect of ablation of the barrel cortex Behav Brain Res44 87 plusmn 99

BELFORD GR and HP KILLACKEY (1979a) The develop-ment of vibrissae representation in subcortical trigeminalcenters of the neonatal rat J Comp Neurol 188 63 plusmn 74

BELFORD GR and HP KILLACKEY (1979b) Vibrissarepresentation in subcortical trigeminal centers of theneonatal rat J Comp Neurol 183 305 plusmn 322

BENETT-CLARKE CA NL CHIAIA MF JACQUIN andRW RHOADES (1992) Parvalbumin and calbindinimmunocytochemistry reveal functionally distinct cellgroups and vibrissa-related patterns in the trigeminalbrainstem complex of the adult rat J Comp Neurol 320323 plusmn 338

BERMEJO R D HOUBEN and HP ZEIGLER (1996)Conditioned whisking in the rat Somatosen Motor Res13 225 plusmn 234

BOWDEN REM and ZY MAHRAN (1956) The func-tional significance of the pattern of innervation of themuscle quadratus labii superioris of the rabbit cat andrat J Anat 90 21plusmn 227

BOWER JM DH BEERMAN JM GIBSON GMSHAMBES and W WELKER (1981) Principles of organiza-tion of a cerebro-cerebellar circuit Micromapping theprojections from cerebral (SI) to cerebellar (granule celllayer) tactile areas of rats Brain Behav Evol 18 1plusmn 18

BRUCE LL JG MCHAFFIE and BE STEIN (1987) Theorganization of trigeminotectal and trigeminothalamicneurons in rodents a double-labeling study with fluores-cent dyes J Comp Neurol 262 315plusmn 330

BUZSAKI G (1991) The thalamic clock emergent networkproperties Neuroscience 41 351plusmn 364

CAJAL SR (1911) Histologie du Syst Aacuteeme Nerveux delrsquoHomme et des Vert AcircebrAcirces Maloine Paris

CARVELL GE SA MILLER and DJ SIMONS (1996) Therelationship of vibrissal motor cortex unit activity towhisking in the awake rat Somatosen Motor Res 13115 plusmn 127

CARVELL GE and DJ SIMONS (1986) Somatotopicorganization of the second somatosensory area (SII) inthe cerebral cortex of the mouse Somatosen Res 3213 plusmn 237

CARVELL GE and DJ SIMONS (1987) Thalamic andcorticocortical connections of the second somatic sen-sory area of the mouse J Comp Neurol 265 409 plusmn 427

CARVELL GE and DJ SIMONS (1990) Biometric analy-ses of vibrissal tactile discrimination in the rat J Neurosci10 2638plusmn 2648

CARVELL GE and DJ SIMONS (1995) Task- and subject-related differences in sensorimotor behavior duringactive touch Somatosen Motor Res 12 1plusmn 9

CARVELL GE DJ SIMONS SH LICHTENSTEIN and PBRYANT (1991) Electromyographic activity of mystacialpad musculature during whisking behavior in the ratSomatosen Motor Res 8 159 plusmn 164

CAULLER LJ and BW CONNORS (1994) Synapticphysiology of horizontal afferents to layer 1 in slices ofrat S1 neocortex J Neurosci 14 751plusmn 762

CHAPIN JK M SADEQ and JLU GUISE (1987)Corticocortical connections within the primary somato-sensory cortex of the rat J Comp Neurol 263326 plusmn 346

CHAPIN JK JS SCHNEIDER M NICOLELIS and C-SLIN (1990) A major direct GABAergic pathway fromzona incerta to neocortex Science 248 1553plusmn 1556

CHIAIA NL RW RHOADES CA BENNETT-CLARK SEFISH and HP KILLACKEY (1991a) Thalamic processingof vibrissal information in the rat I Afferent input to themedial ventral posterior and posterior nuclei J CompNeurol 314 201plusmn 216

CHIAIA NL RW RHOADES SE FISH and HPKILLACKEY (1991b) Thalamic processing of vibrissalinformation in the rat II Morphological and functionalproperties of medial ventral posterior nucleus andposterior nucleus neurons J Comp Neurol 314217plusmn 236

CLARKE WB and D BOWSHER (1962) Terminal distribu-tion of primary afferent trigeminal fibers in the rat ExpNeurol 6 372 plusmn 383

DE BELLESCIZE H (1932) La reception synchrone OndeElectrique 11 230 plusmn 240

DENIAU JM H KITA and ST KITAI (1992) Patterns oftermination of cerebellar and basal ganglia efferents inthe rat thalamus Strictly segregated and partly over-lapping projections Neurosci Lett 144 202 plusmn 206

DESCH AringENES M J BOURASSA and A PARENT (1996)Striatal and cortical projections of single neurons fromthe central lateral thalamic nucleus in the rat Neuros-cience 72 679 plusmn 687

DIAMOND ME (1995) Somatosensory thalamus of therat In Cerebral Cortex the Barrel Cortex of Rodents EGJONES and A PETERS eds pp 189 plusmn 209 Plenum PressNew York

DIAMOND ME MJ BUDWAY M ARMSTRONG-JAMESand FF EBNER (1992) Somatic sensory responses in therostral sector or the posterior group (POm) and in theventral posterior medial nucleus (VPM) of the ratthalamus dependence on the barrel field cortex J CompNeurol 319 66plusmn 84

DINSE HR F SPENGLER K KOPECZ and G SCHONER

(1992) Evoked oscillatory cortical responses are dynam-ically coupled to peripheral stimuli Theoret Neurosci 3579 plusmn 582

DONOGHUE JP KL KERMAN and FF EBNER (1979)Evidence of two organizational plans within the somaticsensory-motor cortex of the rat J Comp Neurol 183647plusmn 664

DONOGHUE JP and ST KITAI (1981) A collateralpathway to the neostratum from corticofugal neurons ofthe rat sensory-motor cortex an intracellular HRPstudy J Comp Neurol 201 1plusmn 13

DONOGHUE JP and JN SANES (1988) Organization ofadult motor cortex representation patterns followingneonatal forelimb nerve injury in rats J Neurosci 83221plusmn 3232

DORFL J (1982) The musculature of the mystacialvibrissae of the white mouse J Anat 135 147plusmn 154

DORFL J (1985) The innervation of the mystacial regionof the white mouse A topographical study J Anat 142173 plusmn 184

DRAGER UC and DH HUBEL (1976) Topography ofvisual and somatosensory projections to the mousesuperior colliculus J Neurophysiol 39 91plusmn 101

DURHAM D and TA WOOLSEY (1977) Barrels andcolumnar cortical organization evidence from 2-deox-yglucose (2-DG) experiments Brain Res 137169 plusmn 174


ERZURUMLU RS and HP KILLACKEY (1980) Dience-phalic projections of the subnucleus interpolaris of thebrainstem trigeminal complex in the rat Neuroscience 51891plusmn 1901

ERZURUMLU RS and HP KILLACKEY (1979) Efferentconnections of the brainstem trigeminal complex withthe facial nucleus of the rat J Comp Neurol 18875plusmn 86

FABRI M and H BURTON (1991a) Ipsilateral corticalconnections of primary somatic sensory cortex in rats JComp Neurol 311 405 plusmn 424

FABRI M and H BURTON (1991b) Topography connec-tions between primary somatosensory cortex and poste-rior complex in rat a multiple fluorescent tracer studyBrain Res 538 351plusmn 357

FAULL RLM and WR MEHLER (1978) The cells oforigin of nigrotectal nigrothalamic and nigrostriatalprojections in the rat Neuroscience 3 989 plusmn 1002

FEE MS PP MITRA and D KLEINFELD (1996) Auto-matic sorting of multiple unit neuronal signals in thepresence of anisotropic and non-Gaussian variability JNeurosci Meth 69 175 plusmn 188

FEE MS PP MITRA and D KLEINFELD (1997) Centralversus peripheral determinates of patterned spike activ-ity in rat vibrissa cortex during whisking J Neurophysiol78 1144plusmn 1149

GERFEN CR WA STAINES GW ARBUTHNOTT andHC FIBIGER (1982) Crossed connections of the sub-stantia nigra in the rat J Comp Neurobiol 207283 plusmn 303

GETTING PA (1989) Emerging principles governing theoperation of neural networks Ann Rev Neurosci 12185 plusmn 204

GIOANNI Y and M LAM ARCHE (1985) A reappraisal of ratmotor cortex organization by intracortical microstimula-tion Brain Res 344 49 plusmn 61

GOLDREICH D TL PRIGG GE CARVELL and DJSIMONS (1997) Neurophysiological recordings during acomputer-automated vibrissal tactile discriminationtask Soc Neurosci Abstr 23 23

GUIC-ROBLES E WM JENKINS and H BRAVO (1992)Vibrissal roughness discrimination is barrel cortex-dependent Behav Brain Res 48 145 plusmn 152

GUIC-ROBLES E C VALDIVIESO and G GUAJARDO

(1989) Rats can learn a roughness discrimination usingonly their vibrissal system Behav Brain Res 31285 plusmn 289

HALL RD and EP LINDHOLM (1974) Organization ofmotor and somatosensory neocortex in the albino ratBrain Res 66 23 plusmn 38

HAROIAN AJ LC MASSOPUST and PA YOUNG (1981)Cerebellothalamic projections in the rat an autoradio-graphic and degeneration study J Comp Neurol 197217plusmn 236

HOOGLAND PV E WELKER and H VAN DER LOOS

(1987) Organization of the projections from barrelcortex to thalamus in mice studied with Phaseolusvulgaris-leucoagglutinin and HRP Exp Brain Res 6873plusmn 87

HUERTA M A FRANKFURTER and J HARTING (1983)Studies of the principal sensory and spinal trigeminalnuclei of the rat projections to the superior colliculusinferior olive and cerebellum J Comp Neurol 220147plusmn 167

HUTSON KA and RB MASTERTON (1986) The sensorycontribution of a single vibrissarsquos cortical barrel JNeurophysiol 56 1196plusmn 1223

ISOKAWA-AKESSON M and BR KOM ISARUK (1987)Difference in projections to the lateral and medialfacial nucleus anatomically separate pathways forrhythmical vibrissa movement in rats Exp Brain Res 65385 plusmn 398

IZRAELI R and LL PORTER (1995) Vibrissal motorcortex in the rat connections with the barrel field ExpBrain Res 104 41plusmn 54

JACQUIN M M BARCIA and RW RHOADES (1989)Structure plusmn function relationships in rat brainstem sub-nucleus interpolaris IV Projection neurons J CompNeurol 282 45plusmn 62

JACQUIN MF NL CHIAIA JH HARING and RWRHOADES (1990a) Intersubnuclear connections withinthe rat trigeminal brainstem complex Somatosen MotorRes 7 399 plusmn 420

JACQUIN MF MR WIEGAND and WE RENEHAN

(1990b) Structure plusmn function relationships in rat brainstem subnucleus interpolaris VIII Cortical inputs JNeurophysiol 64 3 plusmn 27

KAM Z HB SHORE and G FEHER (1975) Simpleschemes for measuring autocorrelation functions RevScientif Instr 46 269 plusmn 277

KAMEN G and GE CALDWELL (1996) Physiology andinterpretation of the electromyogram J Clin Neu-rophysiol 13 366 plusmn 384

KASSEL J (1980) Superior colliculus projections to tactileareas of rat verebellar hemispheres Brain Res 202291plusmn 305

KELLER A (1993) Intrinsic synaptic organization of themotor cortex Cerebral Cortex 3 430 plusmn 441

KELLER A (1995) Synaptic organization of the barrelcortex In Cerebral Cortex the Barrel Cortex of RodentsEG JONES and A PETERS eds pp 221plusmn 262 PlenumPress New York

KENNEDY TT RJ GRIM M and AL TOWE (1966) Therole of cerebral cortex in evoked somatosensory activityin cat cerebellum Exp Neurol 14 13 plusmn 32

KILLACKEY H and R ERZURUMLU (1981) Trigeminalprojections to the superior colliculus of the rat J CompNeurol 201 221plusmn 242

KILLACKEY HP M JACQUIN and RW RHOADES (1990)Development of somatosensory system structures InDevelopment of Sensory Systems in Mammals JR COLE-

MAN ed pp 403 plusmn 429 John Wiley New YorkKIM U E GREGORY and WC HALL (1992) Pathway

from the zona incerta to the superior colliculus in the ratJ Comp Anat 321 555 plusmn 575

KLEINFELD D and KR DELANEY (1996) Distributedrepresentation of vibrissa movement in the upper layersof somatosensory cortex revealed with voltage sensitivedyes J Comp Neurol 375 89 plusmn 108

KLEINFELD D and H SOMPOLINSKY (1988) Associativeneural network model for the generation of temporalpatterns theory and application to central patterngenerators Biophys J 54 1039plusmn 1051

KOCH C and T POGGIO (1992) Multiplying withsynapses and neurons In Single Neuron Computation TMCKENNA J DAVIS and SF ZORNETZER eds pp315 plusmn 345 Academic Press Boston

KOLMAC CI BD POWER and J MITROFANIS (1998)Patterns of connections between zona incerta andbrainstem in rats J Comp Neurol 396 544plusmn 555

KORALEK K KF JENSEN and HP KILLACKEY (1988)Evidence for two complementary patterns of thalamicinput to the rat somatosensory cortex Brain Res 463346 plusmn 351

LEE HS RJ KOSINSKI and GA MIHAILOFF (1989)Collateral branches of cerebellopontine axons reach thethalamus superior colliculus or inferior olive a dou-ble-fluorescence and combined fluorescence-horserad-ish peroxidase study in the rat Neuroscience 28725 plusmn 734

L Acircevesque M A CHARARA S GAGNON A PARENT andM DESCH AtildeENES (1996) Corticostriatal projections fromlayer V cells in rat are collaterals of long-range corticofu-gal axons Brain Res 709 311plusmn 315


LUND RD and KE WEBSTER (1967) Thalamic afferentsfrom the spinal cord and trigeminal nuclei J CompNeurol 130 313 plusmn 328

MA PKM and TA WOOLSEY (1984) Cytoarchi-tectural correlates of vibrissae in the medullarytrigeminal complex of the mouse Brain Res 306374 plusmn 379

MANTLE-ST JOHN LA and DJ TRACEY (1987) Somato-sensory nuclei in the brainstem of the rat independentprojections to the thalamus and cerebellum J CompNeurol 255 259 plusmn 271

MARTIN MR and D LODGE (1977) Morphology of thefacial nucleus of the rat Brain Res 38 206 plusmn 210

MCCORMICK DA and T BAL (1997) Sleep and arousalthalamocortical mechanisms Ann Rev Neurosci 20185 plusmn 215

MERCIER BE CR LEGG and M GLICKSTEIN (1990)Basal ganglia and cerebellum receive different somato-sensory information in rats Proc Natl Acad Sci USA 874388plusmn 4392

MIDDLETON FA and PL STRICK (1997) Cerebellaroutput channels Int Rev Neurobiol 41 61plusmn 82

MIYASHITA E A KELLER and H ASANUMA (1994)Inputplusmn output organization of the rat vibrissal motorcortex Exp Brain Res 99 223 plusmn 232

MIYASHITA E and M SHIGEMI (1995) The superiorcolliculus relays signals descending from the vibrissalmotor cortex to the facial nerve nucleus in the ratNeurosci Lett 195 69 plusmn 71

MOGOSEANU D AD SM ITH and JP BOLAM (1994)Monosynapatic innervation of facial motoneurones byneurones of the parvicellular reticular formation ExpBrain Res 101 427 plusmn 438

MORISETTE J and JM BOWER (1996) Contribution ofsomatosensory cortex to responses in the rat cerebellumgranule cell layer following peripheral tactile stimula-tion Exp Brain Res 109 240 plusmn 250

MOXON KA RS MARKOWITZ and JK CHAPIN (1998)Neural ensemble activity during a whisker discrimina-tion task sensory modulation at different levels of thetrigeminal system Soc Neurosci Abstr 24 133

NAKAJIMA S Y KOMATSU and K TOYAMA (1988)Synaptic action of layer 1 fibers on cells in cat striatecortex Brain Res 457 176 plusmn 180

NEAFSEY EJ EL BOLD G HAAS KM HURLEY-GIUSG QUIRK CF SIEVERT and RR TERREBERRY

(1986) The organization of the rat motor cortex amicrostimulation mapping study Brain Res Rev 1177plusmn 96

NICOLELIS MA AA GHAZANFAR BM FAGGIN SVOTAW and LM OLIVEIRA (1997) Reconstructing theengram simultaneous multisite many single neuronrecordings Neuron 18 529plusmn 537

NICOLELIS MAL LA BACCALA RCS LINand JK CHAPIN (1995) Sensorimotor encoding bysynchronous neural ensemble activity at multiplelevels of the somatosensory system Science 2681353plusmn 1358

NICOLELIS MAL JK CHAPIN and RCS LIN (1992)Somatotopic maps within the zona incerta relay parallelGABAergic somatosensory pathways to the neocortexsuperior colliculus and brainstem Brain Res 577134 plusmn 141

NICOLELIS MAL RCS LIN DJ WOODWARD and JKCHAPIN (1993) Dynamic and distributed properties ofmany-neuron ensembles in the ventral posterior medialthalamus of awake rats Proc Natl Acad Sci USA 902212plusmn 2216

NIEMI-JUNKOLA UJ and GWM WESTBY (1998) Spatialvariation in the effects of inactivation of substantia nigraon neuronal activity in rat superior colliculus NeurosciLett 241 175 plusmn 179

NOTHIAS F M PESCHANSKI and J-M BESSON (1988)Somatotopic reciprocal connections between the soma-tosensory cortex and thalamic Po nucleus in rat BrainRes 447 169 plusmn 174

NOZAKI S A IRIKI and Y NAKAM URA (1986) Location ofcentral rhythm generator involved in cortically inducedrhythmical masticatory jaw-opening movement in theguinea pig J Neurophysiol 55 806 plusmn 825

PAPEZ JW (1927) Subdivisions of the facial nucleus JComp Neurol 42 159 plusmn 191

POWER BD CI KOLMAC and J MITROFANIS (1999) Evi-dence for a large projection from the zona incerta to thedorsal thalamus J Comp Neurol 404 554 plusmn 565

RICE FL and J ARVIDSSON (1991) Central projections ofprimary sensory neurons innervating different parts ofthe vibrissae follicles and intervibrissal skin on themystacial pad of the rat J Comp Neurol 309 1plusmn 16

RICE FL BT FUNDIN K PFALLER and J ARVIDSSON

(1994) The innervation of the mystacial pad in the adultrat studied by anterograde transport of HRP conjugatesExp Brain Res 99 233 plusmn 246

RICE FL A MANCE and BL MUNGER (1986) Acomparative light microscopic analysis of the sensoryinnervation of the mystacial pad I Innervation ofvibrissal follicleplusmn sinus complexes J Comp Neurol 252154 plusmn 174

SACHDEV RNS EW JENKINSON and FF EBNER (1998)Response properties of barrel field neurons in awakebehaving rat Soc Neurosci Abstr 24 633

SANES JN S SUNER and JP DONAHUE (1990) Dynamicorganization of primary motor cortex output to targetmuscles in adult rats II Long-term patterns of reorgani-zation following motor or mixed peripheral nervelesions Exp Brain Res 79 479 plusmn 491

SEMBA K and BR KOMISARUK (1984) Neural substratesof two different rhythmical vibrissal movements in therat Neuroscience 12 761plusmn 774

SHAMBES GM JM GIBSON and W WELKER (1978)Fractured somatotopy in granule cell tactile areas of ratcerebellar hemispheres revealed by micromapping BrainBehav Evol 15 94plusmn 140

SIMONS DJ (1978) Response properties of vibrissal unitsin rat S1 somatosensory neocortex J Neurophysiol 41798 plusmn 820

SIMONS DJ GE CARVELL and SH LICHTENSTEIN

(1990) Responses of rat trigeminal ganglion neurons tomovements of vibrissae in different directions SomatosenMotor Res 7 47plusmn 65

SIMONS DJ and JA HARTINGS (1998) Thalamic relay ofafferent responses to 1- to 12-Hz whisker stimulation inthe rat J Neurophysiol 80 1016plusmn 1019

SMITH RL (1973) The ascending fiber projections fromthe principal sensory trigeminal nucleus in the rat JComp Neurol 148 423 plusmn 446

STEIN BE B MAGALHAES-CASTRO and L KRUGER

(1975) Superior colliculus visuotopic plusmn somatotopicoverlap Science 189 224plusmn 225

STEINDLER DA (1985) Trigeminocerebellar trigemino-tectal and trigeminothalamic projections a doubleretrograde axonal tracing study in the mouse J CompNeurol 237 155 plusmn 175

STERIADE M DA MCCORM ICK and TJ SEJNOWSKI

(1993) Thalamocortical oscillations in the sleeping andaroused brain Science 262 679 plusmn 685

SUGITANI M J YANO T SUGAI and H OOYAMA (1990)Somatopic organization and columnar structure of thevibrissae representation in the rat ventrobasal complexExp Brain Res 81 346 plusmn 351

SWENSON RS RJ KOSINSKI and AJ CASTRO (1984)Topography of spinal dorsal column nuclear and spinaltrigeminal projections to the pontine gray in rats J CompNeurol 222 301plusmn 311


TOKUNO H M TAKADA Y IKAI and N MIZUNO (1994)Direct projections from the deep layers of the superiorcolliculus to the subthalamic nucleus in the rat BrainRes 639 156 plusmn 160

TORVIK A (1956) Afferent connections to the sensorytrigeminal nuclei the nucleus of the solitary tract andadjacent structures An experimental study in the rat JComp Neurol 106 51plusmn 132

VAN DER LOOS H (1976) Barreloids on the mousesomatosensory thalamus Neurosci Lett 2 1plusmn 6

VINCENT SB (1912) The function of the vibrissae in thebehavior of the white rat Behav Monogr 1 7plusmn 81

VINCENT SB (1913) The tactile hair of the white rat JComp Neurol 23 1plusmn 23

WATSON CRR and RC SWITZER III (1978) Trigeminalprojections to cerebellar tactile areas in the rat Originmainly from n interpolaris and n principalis NeurosciLett 10 77plusmn 82

WEISS DS and A KELLER (1994) Specific patterns ofintrinsic connections between representation zones inthe rat motor cortex Cerebral Cortex 4 205 plusmn 214

WELKER C (1976) Receptive fields of barrels in thesomatosensory neocortex of the rat J Comp Neurol 166173 plusmn 190

WELKER E PV HOOGLAND and H VAN DER LOOS

(1988) Organization of feedback and feedforward pro-jections of the barrel cortex A PHA-L study in themouse Exp Brain Res 73 411plusmn 435

WELKER WI (1964) Analysis of sniffing of the albino ratBehaviour 12 223 plusmn 244

WELSH JP EJ LANG I SUGLHARA and R LLINAS

(1995) Dynamic organization of motor control withinthe olivocerebellar system Nature 374 453plusmn 457

WESTBY GW C COLLINSON and P DEAN (1993)Excitatory drive from deep cerebellar neurons to the

superior colliculus in the rat an electrophysiologicalmapping study Eur J Neurosci 5 1378plusmn 1388

WESTBY GW C COLLINSON P REDGRAVE and P DEAN

(1994) Opposing excitatory and inhibitory influencesfrom the cerebellum and basal ganglia converge on thesuperior colliculus an electrophysiological investigationin the rat Eur J Neurosci 6 1335plusmn 1342

WHITE EL and RA DEAMICIS (1977) Afferent andefferent projections of the region in mouse Sml cortexwhich contain the posteromedial barrel subfield J CompNeurol 175 455 plusmn 482

WIESENDANGER R and M W IESENDANGER (1982a) Thecorticopontine system in the rat I Mapping of cortico-pontine neurons J Comp Neurol 208 215 plusmn 226

WIESENDANGER R and M W IESENDANGER (1982b) Thecorticopontine system in the rat II The projectionpattern J Comp Neurol 208 227plusmn 238

WILLIAM S MN DS ZAHM and MF JACQUIN (1994)Differential foci and synaptic organization of the princi-pal and spinal trigeminal projections to the thalamus inthe rat Eur J Neurosci 6 429 plusmn 453

WINESKI LE (1983) Movements of the cranial vibrissaein the golden hamster (Mesocritus auratus) J Zool (Lond )200 261plusmn 280

WINESKI LE (1985) Facial morphology and vibrissalmovement in the golden hamster J Morphol 183199 plusmn 217

WISE SP and EG JONES (1977) Cells of origin andterminal distribution of descending projections of therat somatic sensory cortex J Comp Neurol 175129 plusmn 158

WOOLSEY TA C WELKER and RH SCHWARTZ (1974)Comparative anatomical studies of the SmI face cortexwith special reference to the occurrence of ordf barrelsordm inlayer IV J Comp Neurol 164 79plusmn 94


whisking at the level of sensory and motor cortex (v)the nature of vibrissa control along the motorpathway and (vi) mechanisms for computing the

position of the vibrissae upon their contact with anobject

Mechanics of whisking

The face of the rat has a muscular thickeningdenoted the mystacial pad that contains upwardsof 40 long hairs referred to as vibrissae that arenominally arranged in a Manhattan-like grid withfive rows Each vibrissa sits in a follicle that con-tains sensory nerve endings that originate from abranch of the trigeminal nerve (infraorbital branchof the 5th cranial nerve) (Dorfl 1985 Rice et al1986) The follicle is attached to the mystacial padnear the surface of the skin This attachment servesas a pivot and the follicle is propelled by contrac-tion of a muscular sling that is innervated by abranch of the facial nerve (7th cranial nerve) (Fig1b) (Dorfl 1985 Wineski 1985 Rice et al1994) One feature of this arrangement is thatprotraction is active while retraction is passiveThus the frequency of whisking is ultimately lim-ited by the viscoelastic properties of the mystacialpad A second feature of this arrangement is thatactivation of the sensory afferents is expected todepend primarily on a change in the position ofthe vibrissa relative to that of the follicle as mayoccur upon contact of the vibrissae with an exter-nal object or during periods of rapid accelerationof the vibrissae in the absence of contact

The motion of the vibrissae is governed bymuscles that move the entire mystacial pad as wellas the muscles that control each follicle (Wineski1985) Despite the potential for individual vibris-sae to move independent of each other the vibris-sae are observed to move largely as a single unitduring a multitude of exploratory behaviors (Vin-cent 1912 Wineski 1983) Further the vibrissaemove with bilateral symmetry (Fig 1a) Thuswhisking may be described solely in terms of singleangle as a function of time at least to the extentthat translational motion of the mystacial pad canbe ignored The change in position is described interms of a rapidly changing phase within eachwhisk cycle and a slowly changing set-point andmaximum amplitude (Fig 1c) We will show laterthat this description allows the position of thevibrissae to be measured via the electromyogram(EMG) at a single location in the mystacial pad

A final issue concerns the lack of proprioceptivefeedback of vibrissa position In skeletal joints theextent of muscle contraction is coded by the innerva-tion of specialized muscle fibers known as spindlefibers that provide feedback on the actual motion ofthe muscles as part of a reflex loop However there isno evidence for spindle fibers in facial musculature(Bowden and Mahran 1956) Consistent with thisgeneral result unpublished anatomical studies (F LRice personal communication) have confirmed theabsence of spindle fibers throughout the musculature


FIGURE 1 The nature and control of exploratory whiskingin the rat (a) Consecutive video frames of the head of a ratas it produces large exploratory whisks in search of a foodtube The vibrissae were highlighted by strob illuminationThe time between frames is 33 ms (b) The sling-likearrangement of the musculature surrounding each folliclein the mystacial pad The lower structure is one of twosensory nerves (adapted from Dorfl (1982)) (c) Diagramof the cyclic motion of the vibrissae during exploratorywhisking The animal whisks with frequency n D The set-point and maximum amplitude of the rhythmic whiskingvary slowly on the timescale of each whisk cycle



control


Hindbrain loop



Midbrain loop







colliculus

















whisking












ordf twitchingordm



























whisking













loop



















n R = n D (Eqn 1)



















Concluding remarks









Acknowledgements


Notes











References






































































































































































control


Hindbrain loop



Midbrain loop







colliculus

















whisking












ordf twitchingordm



























whisking













loop



















n R = n D (Eqn 1)



















Concluding remarks









Acknowledgements


Notes











References






































































































































































colliculus

















whisking












ordf twitchingordm



























whisking













loop



















n R = n D (Eqn 1)



















Concluding remarks









Acknowledgements


Notes











References











































































































































































whisking












ordf twitchingordm



























whisking













loop



















n R = n D (Eqn 1)



















Concluding remarks









Acknowledgements


Notes











References






































































































































































ordf twitchingordm



























whisking













loop



















n R = n D (Eqn 1)



















Concluding remarks









Acknowledgements


Notes











References
























































































































































































whisking













loop



















n R = n D (Eqn 1)



















Concluding remarks









Acknowledgements


Notes











References






































































































































































loop



















n R = n D (Eqn 1)



















Concluding remarks









Acknowledgements


Notes











References

















































































































































































n R = n D (Eqn 1)



















Concluding remarks









Acknowledgements


Notes











References














































































































































































Concluding remarks









Acknowledgements


Notes











References





































































































































































Acknowledgements


Notes











References




































































































































































invited review anatomical loops and their electrical ... · correspondence: david kleinfeld,...

Documents