Integrated technology for evaluation ofbrain function and neural plasticity

Paolo M. Rossini, MDa,b,Gloria Dal Forno, MD, PhDb,c,*

aDepartment of Clinical Neuroscience, Hospital Fatebenefratelli,

Isola Tiberina 39, 00186-Rome, ItalybDepartment of Neurology, University Campus Bio-Medico,

Via Longoni 83, Rome, 00155-ItalycDepartment of Neurology, The Johns Hopkins University School

of Medicine, Baltimore, MD, USA

It traditionally has been taught that adult brain has no signicant abilityfor self-repair or reorganization after injuries that cause neuronal death. It iscommon in clinical practice, however, to see slow but consistent recoveryover weeks and months after lesions of apparently stabilized neurologicdecits, such as stroke [1]. Degrees of spontaneous recovery range widely,even in light of similar acute lesions and clinical pictures.

Several explanations have been given for the recovery of lost function. Instroke syndromes, reabsorption of perilesional edema and interindividualvariability of perfusion patterns and arterial collaterals have been proved toplay major roles. Additional and more general mechanisms for humancentral nervous system (CNS) functional recovery, which, despite extensiveanimal research [2,3], had been elusive, now are being understood anddescribed thanks to new techniques for their functional study in vivo [4].Multiple representations of the same function in the cortex (eg, motorcontrol), presence and amount of alternative neural routes (eg, ipsilateralcorticospinal bers) [5], unmasking of functionally silent connections, andchanges in synaptic ecacy and remodeling all are important contributorsto the plasticity of the human CNS. A better understanding of the

Phys Med Rehabil Clin N Am

15 (2004) 263306mechanisms underlying recovery (or deterioration) of function after a CNSlesion and the mechanisms leading to maladaptive or unfavorable outcomes

* Corresponding author. Cattedra di Neurologia, University Campus Bio-Medico, Via

Longoni 83, Rome, 00155-Italy.

E-mail address: G.Dalforno@unicampus.it (G. Dal Forno).

1047-9651/03/$see front matter 2004 Elsevier Inc. All rights reserved.doi:10.1016/S1047-9651(03)00124-4

is essential for directing specic and eective rehabilitative strategies andavoiding potentially harmful interventions.

This article reviews the anatomic and physiologic aspects of the CNS thatlikely play a major role in postlesional functional improvement, with specialemphasis on the sensorimotor system. Methods that allow for in vivoevaluation of plastic phenomena are reviewed.

Mechanisms of plasticity of the central nervous system

Experimental research has shown that mature CNS does have a degree ofcapacity for self-repair and reorganization after injury, even though thedenite clinical demonstration of the linkage between functional recovery andplastic reorganization still is lacking [5]. The term plasticity describes thecapacity for pliancy andmalleability.When applied to a dynamic and biologicsystem, such as brain function and behavior, the concept of plasticity usuallyincludes the potential for change and all the mechanisms of self-repair or ofreorganization of neural connections. Plastic phenomena probably are at thebasis of learning and of damage repair; cortical maps can be modied bysensory input, experience, and learning [68]. Cortical representation areasprobably undergo continuous transient changes during routine life experi-ences, in response to repeated stimuli, movement patterns, and cognitive tasks[9]. Transient enlargement of cortical representation areas over a few days hasbeen shown in response to the repeated performance of a skilled movementpattern, such as the learning of a piano exercise [10]. These changes are likelyto become stable, depending on the duration of the exposure to a stimulusor motor pattern, as exemplied by the permanent, asymmetric enlargementof the cortical representation area of the left ngers in string players [11].

The physiology at the basis of the rapid plastic changes of the CNSprobably includes many cellular and anatomic phenomena. The true ana-tomic distribution of a neural network is much larger than the area of its usualfunctional inuence, as exemplied by the multiple representations of eachmuscle and joint area in the motor cortex (see later). Modications of thesynaptic ecacy within neuronal networks [12] have been favored by manyinvestigators as underlying learning and memory and some of the corticalplasticity related to acquisition and recovery of sensorimotor function.

The functional status and distribution of a network probably depends onthe balance between excitation and inhibition, with some areas being keptsilent, or masked, by a mechanism of active tonic inhibition mediatedthrough GABAergic input. This active inhibition can be altered or removed,causing a rapid change in size or distribution of the functional network,a process called unmasking [13]. Modications of the neuron membrane

+

264 P.M. Rossini, G. Dal Forno / Phys Med Rehabil Clin N Am 15 (2004) 263306excitability, through changes in Na channel activation, possibly mediatedby protein kinase C activation, also may play a role [14].

Other relatively fast processes that most likely play a pivotal role in thefunctional strengthening or weakening of existing synapses are long-term

potentiation (LTP) and long-term inhibition (LTI) [15]. LTP and LTI areneurophysiologic mechanisms by which information is stored in the CNS.LTI probably is based on Na+ or Ca2+ channels at least for short-termchanges, whereas LTP-like mechanisms include glutamate-dependent N-methyl-D-aspartate receptor activation for long-term changes, a process thatcan be antagonized by c-aminobutyric acid (GABA) inhibition [16]. Eversince it was discovered in the hippocampus, LTP has been regarded as theprototypic mechanism of modied synaptic ecacy [17]. LTP can occur withdierent mechanisms, such as modication of the input to a postsynaptic cell,the concomitant synchronous input to the synapsis from another cell, orpostsynaptic depolarization [18]. Converging inputs from various sources,including corticocortical and subcorticocortical connections, could interactand lead to the reshaping of cortical input-output somatotopy [1923].

The release of neurotransmitters from diuse projection systems, such asnorepinephrine synthesized by the locus caeruleus cells, dopamine by thelateral tegmentum, and serotonin by the raphe nuclei, also is likely tomodulate synaptic ecacy [24,25]. The role of nitric oxide as a potentialmodulator of cortical plasticity also is under investigation [26].

A fourth process likely to play a role in CNS plasticity, althoughconceivably over a longer time frame, is neuronal sprouting and formation ofnew synaptic connections [27]. These changes at a cellular level probably areinuenced strongly by release of local neurotransmitter and neurotrophicfactors and synaptic protein synthesis [28]. In particular, dendrites anddendritic spines, the main site for synaptic connections, undergo continuousremodeling [29], a process most likely also modulated by interaction withneighboring astrocytes [30]. The elements that govern local neurotrophicfactor release probably include many physiologic and pathologic stimuli. It iswell documented that ischemia is a strong inducer of gene expression in thebrain [31,32], and many growth factors can reduce infarct size inexperimental models [33,34]. In experimentally induced stroke, placementin an enriched environment, with the possibility of performing variousactivities and interactions, resulted in better recovery compared withstandard laboratory environments [35]. These ndings may be related todirect stimulation of trophic factor release through environmental stimula-tion, a mechanism probably also operating in normal nonpathologicsituations, such as experience-based or use-dependent plasticity (Fig. 1).

In addition to cellular and synaptic mechanisms, the CNS functionalanatomy seems to be organized, even through a certain degree of relativeredundancy, so that damage at least in part can be compensated functionally[36]. All damaging and restorative mechanisms are inuenced strongly by theremote eects of the loss of excitatory/inhibitorymodulation from a damagedarea onto adjacent or distant brain centers connected via corticocortical and

265P.M. Rossini, G. Dal Forno / Phys Med Rehabil Clin N Am 15 (2004) 263306transcallosal projections, a phenomenon called diaschisis. Corticocorticalinputs to the primary motor cortex (MI) are dominated by inputs from thesupplementary motor area, premotor cortex and primary sensory cortices 1,

2, and 3, where spatiotemporal maps allow the integration of theproprioceptive, tactile, and visual cues necessary for manual actions [37].Voluntary motor activity in humans requires sensory feedback from themoving part [38,39]. After a stroke, patients with poor improvement of handmotor control show severe metabolic depression in the thalamus [40,41].Together with the output for movement, simultaneous discharges to primarysomatosensory cortex (SI) are red by MI, probably providing the eerentcopy of the motor program with which the sensory feedback should bematched [4244]. An appropriate sensory feedback from the paretic handseems fundamental for long-term motor recovery and neuronal reorganiza-tion beyond the limits o f the usual sensorimotor areas [45].

In healthy humans, the organization of the sensorimotor areas issymmetric between the two hemispheres, and this has been shown withdierent methods of functional brain imaging, particularly for hand control.When a hemispheric lesion occurs, adjacent or distant but functionallyconnected neuronal aggregates can take over progressively for the lostneurons [46,47], and this reorganization is expected to modify interhemi-spheric symmetry in terms of absolute surfaces, number of recruitedneurons, and spatial coordinates. Symmetry can be used as an important

Fig. 1. Schematic diagram of a possible mechanism of neuronal plasticity. (A) The release of

neurotransmitter at the synaptic junction during normal neurotransmission. (B) Augmented

release of neurotransmitter during experience-related, short-term learning causes a transient

increase in the activation of the postsynaptic neuron. (C) The repeated or reinforced stimulation

at the synapsis causes neuronal sprouting of the presynaptic neuron and long-term anatomic

changes, which lead to long-lasting increase in the activation of the postsynaptic neuron.

266 P.M. Rossini, G. Dal Forno / Phys Med Rehabil Clin N Am 15 (2004) 263306parameter to study brain lesions.The motor system seems to be the most obvious system to study as a model

for basic plasticity phenomena because it controls readily evident human

behaviors, yet it also is complex enough to be representative of many otherfunctional systems. In the next section, the neurophysiology of sensorimotorfunction, particularly for hand and arm control, is reviewed in more detail.

Physiology of sensorimotor brain areas and of related plasticity

Sensory perception and its role in motor control involves large brainareas, including primary somatosensory, visual, and motor cortices andsecondary sensory and motor areas. Basal ganglia and thalamic relayssignicantly contribute to motor planning, sensory perception, andsensorimotor integration. Supplementary motor and premotor cortices playa pivotal role in motor preparation and execution, which, on their own, areperformed via corticospinal bers from primary motor cortex, under theparallel control of other descending systems, while cerebellar relaysconstantly monitor motor output and execution (Fig. 2).

The existence of multiple yet discrete eerent microzones and macrozonesfrom primary motor cortex now is accepted as the essential organizationalprinciple of this area. Animal studies showed that the MI is organized sothat a particular movement can be elicited through stimulation of dierentMI regions, often several millimeters apart and separated by nonresponsive

267P.M. Rossini, G. Dal Forno / Phys Med Rehabil Clin N Am 15 (2004) 263306Fig. 2. Motor cortical areas and corticospinal pathways stimulated via transcranial magnetic

stimulation allow for mapping of cortical excitability sites and integrity of the entire motor

pathways.

districts [4857]. In addition, bidirectional projections interconnect motorcortex areas for dierent muscle districts [37,53].

Any movement, in contrast to what traditionally was believed accordingto the labeled-line hypothesis, seems to be controlled not by single groups,but by a whole network of neurons, distributed throughout the MI cortex.In this model, the motor output from overlapping cortical territoriesconverges onto single muscles. Similarly the output from any given corticalsite diverges onto multiple muscles, whereas horizontal intracorticalprojections interconnect subregions within MI [52]. This pattern explainsthe considerable overlap in cortical representation of the entire musculaturein area 4, particularly for hand muscles [58,59].

The multiple representations of each muscle and contiguous joints, theoverlap of outputs to dierent motoneurons, and the large corticalrepresentation of hand and ngers are probably the anatomic substrate forthe extraordinary repertoire of possible movement strategies and thecoordination of action initiated by many muscles at dierent joints [53]. Inother cortical maps, such as SI, the connections are precise, and each nger ismapped in an ordered somatotopic arrangement [60,61]. Functionally, strictcoordination of dierent muscle elds is essential, and the stability of theproximal arm and girdle is necessary for the successful execution of ngermovements. Although movements occur in a three-dimensional space withseveral degrees of freedom, the body surface may be envisioned as a two-dimensional sheet, reproduced as a surface in a point-to-point fashion in SI.Multiple representations underlying dierent motor functions of separatebody parts along with distributed and cooperating neural networks mayoverlap spatially and temporally [62,63], oering exibility in motor learningand progressively substituting a related dysfunctional area more easily thanhighly specialized and unique groups of cells would, being fundamental forfunctional (or maladaptive [see later]) plasticity.

Functional connections between SI and motor cortex also are fundamen-tal for motor control. In addition to direct input to MI from the thalamus,the somatosensory cortex is a major source of input to the motor cortex [64]and is the only type of primary sensory cortex with direct access to MI.Experimental evidence for a separate functional organization within dierentareas of SI is available. Input from deep receptors is projected to areas 3a and2, after a relay in the anterodorsal shell of the ventroposterolateral thalamicnucleus, whereas the cutaneous input, relayed from the central core of theventroposterolateral nucleus, is projected selectively to cortical areas 3b(more densely) and 1 [6468].

In the primate, the representation of the glabrous skin of each digit inarea 3b occupies about 1 to 2 mm2. Digit 1 is more laterally represented,whereas digit 5 is the most medial one [6]. Cells from areas 1 and 2 analyze

268 P.M. Rossini, G. Dal Forno / Phys Med Rehabil Clin N Am 15 (2004) 263306more complex, associative features of a stimulus, such as velocity orpresence of edges, which cross over dierent districts of the receptor surfaceof area 3b. Area 3b extracts quantitative data relative to a stimulus in a given

skin locus [69]. Internal connections between the SI areas are extensive: Area3b projects mainly to areas 1 and 2, whereas area 1 also reciprocally isconnected with areas 2 and 3b.

Technologies for noninvasive functional brain imaging

Dierent imaging techniques now are available for investigating brainfunction. Some measure regional blood ow and metabolic changes linkedwith function-related changes in neuronal ring level. These techniquesinclude positron emission tomography (PET) and functional MRI (fMRI).Other techniqueshigh-resolution electroencephalography (EEG), mag-netoencephalography (MEG), and transcranial magnetic stimulation(TMS)analyze the electromagnetic properties of brain neurons.

PET is a relatively invasive technique involving inhalation or injection ofradioactive substances. Regional cerebral blood ow, a marker of synapticactivity in the structurally intact brain (so-called neurovascular coupling),can be measured using various radioisotopes. The subject lies in the scanner,and radiation detectors, positioned all around the head, detect the anatomiclocation within the subjects brain where the encounter of a positron anda natural electron reciprocally annihilating produce gamma radiations. Thesource of these radiations can be mapped, showing the position of theradioisotope in space (ie, where in the brain the blood preferentially isowing while the subject performs a particular activation task) [70]. Theanatomic location of the activation can be determined by mapping the signalchanges into a standardized coordinate system and superimposed ontoa standardized three-dimensional MRI scan of the human brain [7173].Although initially possible across only a group of subjects, more recentadvances in PET technology have made it possible to detect signicantactivations in individual subjects and to map these changes on the subjectsMRI scans, resulting in substantial improvement in spatial resolution andanatomic registration. The temporal resolution of PET is poor because eachmeasurement of regional cerebral blood ow takes around 90 seconds andcannot be repeated for at least 6 minutes. Because of dosimetry, only a fewscans can be obtained in a single session in a given subject. Other majoradvantages of PET are that radioisotopes can be incorporated in specicligands to receptors that one wants to study [74]. This technique is complex inexecution and expensive requiring a cyclotrone to produce essentially in locothe short decaying radioisotopes needed.

With fMRI, the signal depends on changes in deoxyhemoglobinconcentrations, an endogenous paramagnetic substance, due to the complexinterplay between blood ow, blood volume, and blood oxygen saturation.Neuronal activation secondary to a task increases the need for supply of

269P.M. Rossini, G. Dal Forno / Phys Med Rehabil Clin N Am 15 (2004) 263306oxygen and glucose (although oxygen consumption increases less than theincrease in perfusion). This need results in a relative decrease in localdeoxyhemoglobin, so-called blood oxygen leveldependent signal. Blood

oxygen leveldependent signal is the local change in deoxyhemoglobinconcentration that the MRI scanner detects [75,76]. Although the signalchange is small, this is oset by the acquisition of repeated time series so thatthis technique has sucient sensitivity to detect activations from individualsubjects. The acquisition frames of fMRI depend on the specic collectionsequence used, but can be less than 100 msec for a single scan, which allowsone to perform multislice imaging of the whole brain within a short time.The temporal resolution of fMRI is dictated by the timing of the hemo-dynamic response of the neurovascular coupling phenomenon, about 12 sec-onds. It is nevertheless possible to detect dierences in peak activation timebetween brain regions on the order of 1 to 2 seconds [7779].

PET and fMRI can provide a detailed relationship between function andanatomy and comprehensively map the distributed network subtendinga given motor act. There are, however, intrinsic limitations in PET andfMRI. In particular, the length of the examined epochs necessary to computea statistically signicant activation (seconds to minutes) make it dicult todiscriminate the temporal sequence of a phenomenon and to dierentiateneuronal ring decrease from increase (exciting versus inhibiting net eects).In the case of movement paradigms, the activation directly linked to motorprogramming and execution cannot be distinguished from the sensoryfeedback from the moving parts, and the chronologic relationship(hierarchy) between activated areas cannot be studied [80,81].

TMS [82], a safe, painless, and noninvasive technique, is used increasinglyin investigations of brain plasticity. TMS, through a brief, intense magneticeld, created by a circulating electric current within a coil applied directly tothe scalp, permits the mapping of underlying cortical representation areas(Fig. 3) [8284]. When applied over the scalp regions corresponding to the

270 P.M. Rossini, G. Dal Forno / Phys Med Rehabil Clin N Am 15 (2004) 263306Fig. 3. Transcranial magnetic stimulation device and gure-eight stimulating coil.

motor strip, TMS triggers a transient electromyographic response (motorevoked potentials [MEPs]) in the connected target muscles (Fig. 4) [85].The development of coil stimulators of various shape, size, and orientationallows for preferential excitation of discrete brain regions [8689], and themotor output can be mapped somatotopically, with leg and shouldermuscles located more medially and posteriorly and arm and hand muscleslocated more laterally and anteriorly [90]. Changes in cortical maps that areuse dependent or consequent to a lesion usually show two main character-istics. The rst is the enlargement or restriction of the excitable area, withoutchanges in the amplitude-weighted center of the motor output maps, orcenter of gravity (CoG), possibly due to the recruitment or inhibition ofa fringe of adjacent neurons. The second is the migration of the excitablearea outside the usual boundaries, possibly due to a lesion aecting thebrain district where the hot spot is. This migration may be apparent,due to the activation of a secondary hot spot previously hidden by thepredominant one, or may be real, due to the progressive activation of newsynaptic connections.

TMS also is a powerful tool for studying mechanisms of intracorticalinhibition and excitation [91,92]. With neuroimaging methods based onblood ow, such as PET and fMRI, it is not possible to distinguish whetheractivation of a certain brain area corresponds to the activity of excitatory orinhibitory neuronal networks.

271P.M. Rossini, G. Dal Forno / Phys Med Rehabil Clin N Am 15 (2004) 263306Fig. 4. Hot spot (area with lowest excitability threshold) and surrounding areas of cortical

excitability and schematic grid illustrating peripherally recorded motor evoked potentials, for

a hand muscle, superimposed on a diagram of a normal brain. (From Ferreri F, et al. Motor

cortex excitability in Alzheimers disease: A transcranial magnetic stimulation study. Annals of

Neurology 2003;53:1028; with permission.)

Advanced EEG methods allow one to eliminate the contribution ofvolume currents, to obtain reference free recordings, and to disentanglelocoregional rhythmic or transient neuronal activities produced by tangentialand radially oriented generators from discrete brain regions underlyingthe exploring electrode [93,94]. Another promising computational EEGapproach is the analysis of the coherence of the EEG rhythms (h, a, andb) generated in dierent cortical areas. Coherence analysis studies non-invasively the connectivity of dierent brain regions and how thisconnectivity changes in relation to task activation with a high temporalresolution. Coherence EEG methods and the study of the coherence betweenEEG and EMG signals during voluntary motor activity are promising, andthey might permit a more in-depth study of neuronal plastic changessecondary to either physiologic brain function or to disease states [9597].

MEG is a noninvasive technique able to identify spatially the synchronousring of neurons from restricted cortical areas, in relation to spontaneouscerebral activity or in response to external stimuli. The MEG signal is notinuenced by extracerebral tissue layers overlying brain; this allows one tolocalize and measure the intracellular currents in a shallow brain regionexactly below the recording sensor, without any contributions from volumecurrents (Fig. 5). The MEG signal preferentially records the tangentialcomponent of current dipoles generated in the depth of gyri and sulci. MEGfollows the spatial and temporal evolution of a dipolar generator source,which is modeled as an equivalent current dipole (ECD), able to explain 90%or more of the magnetic eld distribution over the scalp. It has a high timeresolution, in the order of 1 msec or less. As a result of its physical properties,MEG allows a precise three-dimensional localization of the ring neuronalpool whenever it produces a dipolar eld distribution at the scalp level [98].Besides the spatial properties, the strength of ECDs (roughly reecting thenumber of neurons ring synchronously) and their orientation can bemeasured, while response morphology provides indirect information on theunderlying neural circuitries. Decrease or increase of dipole strength can bedue to restriction or enlargement of the responsive area studied, possiblybecause of recruitment of a fringe of neurons surrounding those usually ringin response to the incoming stimulus. These variations can be secondary todynamic phenomena, such as use-dependent modulation of synaptic ecacy,changes of excitatory/inhibitory input from adjacent or remote lesioned brainareas (diaschisis), or changes in the amount of sensory information. The CoGof the responsive area in this model is not modied by plastic reorganization,and the ECD baricenter remains stable. When a brain lesion aects the mainaerents or their target relays in primary somatosensory cortex, recovery oflost function can be achieved if an alternative neural circuitry is activated

272 P.M. Rossini, G. Dal Forno / Phys Med Rehabil Clin N Am 15 (2004) 263306progressively. In this case, the CoG and the ECD baricenter of the responsivearea are shifted in space, and the response morphology is modied.

In the following sections, examples of experimental evidence of the app-licability of these functional neuroimaging techniques are presented (Fig. 6).

Experience-dependent plasticity tracked via neuromagnetic methods

Numerous examples of neural plasticity can be found in physiologicconditions, such as skilled or repeated training, and in instances ofpermanent or transient deaerentation of cerebral cortical areas. In theauditory system, plasticity is an inherent phenomenon, which can be shownbehaviorally throughout life in processes subtending the identication of newvoices or the learning of new languages. Adults and children can learn todistinguish speech sounds that they could not hear before training [99,100].The ring patterns of single auditory neurons and auditory cortex functionare altered as a consequence of training and learning new behaviors [101].The mismatched negativity evoked potential is a powerful tool to showfunctional plasticity in the auditory cortex. The mismatched negativityevoked potential is a cortical response probably originating in nonprimaryauditory areas [102,103]. It normally is obtained when an acoustic change is

Fig. 5. A typical magnetoencephalography (MEG) system in a magnetically shielded room.

Room size is generally approximately 4 4 3 m. The photograph on the right shows thecryogenic container, called a dewar, which contains the MEG sensors or SQUIDs (super-

conducting quantum interference devices), usually mounted in a movable gantry for seated or

horizontal positions. TheMEGhelmet is located at one end of the dewar where the patients head

is positioned. (Helmet MEG system, 150 channels; ITAB-University of Chieti, Chieti, Italy.)

273P.M. Rossini, G. Dal Forno / Phys Med Rehabil Clin N Am 15 (2004) 263306introduced in a repetitive sequence of acoustic stimuli; it is largely attentionindependent and sensitive even to barely perceivable changes. Kraus et al[104] showed that in normal adults the mismatched negativity evokedpotential can be incremented greatly after only a few hours of specic

listening discrimination training directed at the recognition of slightlydierent acoustic stimuli. These neurophysiologic changes, attesting tocortical plastic phenomena occurring in response to training, were longlasting and specic to the training. Simple repeated exposure to varyingstimuli tends instead to cause a decrement in mismatched negativity evokedpotentials over time [105].

The auditory cortical plasticity eects of training also have been shown inchildren with learning disabilities and dyslexia. Dyslexic children often showpoor phonemic awareness, a dysfunction most likely at the basis of theirpoor word recognition skills [106109]. Intensive remedial training andtone-speech discrimination exercises can induce behaviorally measurableimprovement in performance [100,110112]. The brain plasticity changes

Fig. 6. Magnetoencephalography (MEG)/MRI integration. (Top left) Superimposition of all

magnetic traces from left hemisphere, after electrical stimulation of the right median nerve. (Top

middle) Spatial distribution of wave M20 (approximately 20 msec after stimulus) shown by

isoeld contours. (Top right) Spatial coordinates of the activated neuronal source. (Bottom left)

MEG/MRI coordinate system dened by three anatomic landmarks. (Bottom right) Cerebral

activation at 21 msec from stimulation of left and right median nerves as identied by MEG, as

shown in individual MRI scans. (From Rossini PM, Pauri F. Contribution of neuromagnetic

integrated methods in evaluating cerebral mechanisms of plastic reorganization in monohemis-

pheric stroke. Neuroscience News 1999;2:115; with permission.)

274 P.M. Rossini, G. Dal Forno / Phys Med Rehabil Clin N Am 15 (2004) 263306subtending these phenomena have been studied with magnetic sourceimaging, a neuroimaging method that provides real-time spatiotemporalmaps of brain activation by measuring electric currents generated by cortical

neurons during a task. Simos et al [113] showed a long-lasting bihemisphericchange in the patterns of cortical activation in children with dyslexia beforeand after intensive remedial training. When compared with controls,dyslexic children had an abnormal baseline cortical activation pattern inresponse to a pseudoword reading task, showing poor left posterior-superiortemporal gyrus activation and aberrant increased activation on thecorresponding right cortical areas. After intensive remedial trainingresulting in behavioral improvement of reading skills, the dyslexic childrenshowed a measurable normalization of the cortical activation pattern onmagnetic source imaging, similar to the pattern seen in controls.

Cortical plasticity eects of training also have been studied in musicians.MEG studies of auditory stimulation with noise-burst or tones have shownthat the electromagnetic signal generator sources for noise-burst wereseparated and located posteriorly from the sources for tones in musicianswith an absolute pitch, but not in nonmusicians. These results have beenascribed to presence of distinct neural networks in the auditory cortex ofmusicians, which may result from plastic reorganization of the cortexbecause of training [114]. In skilled musicians, the cortical representationareas are enlarged to a degree correlated with age at which musical traininghad begun, with no dierences between musicians with an absolute orrelative pitch [115], suggesting extensive and important plastic corticalrearrangement during the early years of life. The auditory system maturespoorly in the absence of appropriate acoustic stimulation [116], and evenunilateral sensorineural hearing loss, despite the mostly bilateral corticalprojections from the peripheral auditory system, may modify informationprocessing in central auditory pathways [117]. The auditory system retainsa great degree of plasticity despite prolonged periods of deafness, proven bythe successful restoring of cortical stimulation obtained through cochlearimplantation, in children and in adults [118]. In otosclerotic patients studiedbefore and after corrective surgery, a signicant recovery of air-conductedauditory function has been shown. The usual tonotopic organization of theauditory cortex during pure-tone stimulation, which was missing inpresurgical recordings, was reacquired after surgery, together with a signif-icant improvement of acoustic function [119].

Enhanced visual function in deaf individuals has been shown in part,suggesting compensatory enhancement of an intact sensory modality whenanother function is decient, similar to the case of greater tactile or auditoryskills in blind subjects. Studies using event-related potentials have shown thatin deaf subjects, the parietal regions are strongly activated by moving visualstimuli, a phenomenon that occurs only to a small degree in individuals withnormal hearing [120]. This phenomenon suggests reorganization of cortical

275P.M. Rossini, G. Dal Forno / Phys Med Rehabil Clin N Am 15 (2004) 263306areas normally not involved in visual processing or an extension of visualcortical elds outside their normal boundaries. In congenitally deafindividuals, the functional enhancement of the visual sensory modalityseems to relate particularly to peripheral elds, where stimuli cause larger

evoked potentials [121] and faster reaction times in deaf compared withnormal hearing subjects. An fMRI study of visual attention to peripheraleld stimuli showed signicantly greater activation of the motion-sensitivecortical areas in congenitally deaf individuals compared with controls [122].

Similar to conditions of auditory cortical deaerentation, combined event-related potentials and PET studies in blind humans indicated signicantactivation of visual cortical areas in response to auditory stimuli and duringhaptic mental rotation tasks and while reading the Braille alphabet orperforming other tactile discrimination tasks, activities not associated withany cortical activation in individuals with normal vision. Studies with TMSshowed an enlargement in the motor hand area representation in Braillereaders and more so in congenitally blind individuals [123128].

Recruitment of visual cortex for somatosensory processing occurs in blindsubjects but not in subjects with normal vision. This cross-modal plasticitymay contribute to sensory compensation in subjects with early-onsetblindness and may account in part for the superior tactile perceptual abilitiesof blind subjects [129]. Repetitive TMS causes a temporary disruption ofcortical function, and when applied to the visual cortex of subjects with early-onset blindness during identication paradigms of Braille or embossedRoman letters, it induces errors in both tasks, distorting tactile perceptions.Repetitive TMS has no eect on tactile performance in nonvisually impairedindividuals, whereas it is known to disrupt their visual performance [130].

The degree of reorganization occurring in late-onset blindness is less wellcharacterized. The previously mentioned repetitive TMSinduced dysfunc-tion of Braille reading in congenitally and early-onset blind individuals doesnot occur in late-onset blind individuals. Likewise, PET has shown thatcongenitally blind subjects show task-specic PET activation of extrastriatevisual areas and parietal association areas during Braille reading, whereassubjects with postpubertal visual loss show additional activation in theprimary visual cortex alone [130]. Animal models using blind-raisedmonkeys have shown cross-modal activation in extrastriate areas withsomatosensory stimulation, but no response in primary visual cortex. Thedierential activation between late-onset and congenitally blind subjectssuggests the possibility of reciprocal activation by visual imagery in subjectswho have experienced vision early in life [131].

TMS studies of cortical motor output maps in blind workers showed rapidand reversible modulation in relation to activity with a larger motor outputfor the rst dorsal interosseous muscle of the Braille reading hand. Thismodulation appeared greater immediately after their working shift comparedwith after 2 days of rest. Similar changes were not evident in other muscles(abductor digiti minimi) not involved in the Braille reading process [132].

The role for TMS in studying plastic changes of the human motor system

276 P.M. Rossini, G. Dal Forno / Phys Med Rehabil Clin N Am 15 (2004) 263306in the acquisition of new ne motor skills is evidenced further by work byLiepert et al [133]. In this study, a brief training session of synchronousmovements of the ipsilateral thumb and foot induced a transient shift of the

CoG of the hand muscle examined (abductor pollicis brevis), of about 7 mmmedially, toward the foot area. The changes observed, which disappearedapproximately 1 hour after the experiment, were ascribed to corticalmodulation interactions between hand and foot representation areas inmotor cortex. Cortical modulation phenomena may be in part the basis forcommonplace performance improvements in speed and accuracy occurringafter practice of complex motor tasks. Similarly the cortical motor areastargeting the contralateral long nger exor and extensor muscles in subjectslearning a one-handed, ve-nger piano exercise were shown to increase insize and the activation threshold to decrease after only 5 days of training [10].

Focal TMS of the motor cortex evokes isolated movements of the thumb,in the same direction of a thumb movement exercise practiced for severalminutes before stimulation. Substantially smaller eects instead followdirect stimulation of corticofugal axons with transcranial electricalstimulation, pointing to cortex as the site of plasticity. Training establisheda rapid and transient change in the cortical network encoding the kinematicdetails (direction of movement) of the practiced movement [134].

Experience-dependent reorganization of the adult MI may underlie theacquisition and long-term retention of motor skills. fMRI studies showedthat after 4 weeks of training of a specic motor sequence, the extent ofcortex activated by the task is enlarged in trained subjects compared withsubjects who had not practiced the same sequence previously and that thesechanges persisted for several months [135].

Use-dependent aberrant plastic reorganization of brain may represent thesubstrate for some neurologic decits or dysfunctions. Focal dystonia tends tooccur with repetitive synchronous movements of the ngers, a movementpattern that is part of many human activities (eg, instrument playing inmusicians). MEG recordings showed that the mean distances of ECDs in theSI area corresponding to the ngers is smaller for the aected hand of dystonicmusicians comparedwith the nonaected hand in the same subjects orwith thehands of nonmusician controls, suggesting fusion of motor output [136].

Mental imagery of movementsnot evoking a contractionmay increasethe capability of acquiring new motor skills. Such a technique has beenapplied successfully as a training procedure to improve the actualperformance without receiving any feedback about the results. TMS is anexcellent tool to show motor output changes taking place during the ideationof movement. Izumi et al [137] described facilitatory eects of motor imagery,as evidenced by a decrease in the excitability threshold, induced by thinkingabout a movement. Nonmotor mental activity and alerting stimuli per sehave been shown to potentiate the amplitude of MEPs in the target muscles,in strict time relationship with a desynchronization of the background EEGactivity of the stimulated brain areas. Movement imagery can direct a specic

277P.M. Rossini, G. Dal Forno / Phys Med Rehabil Clin N Am 15 (2004) 263306facilitation on the prime-mover muscle involved in a mentally simulatedmovement [44,138]. This process was aected specically and asymmetricallywhen the motor cortex, contralateral to the aected limbs, was stimulated

with focal TMS in early unilateral Parkinsons disease patients [139],suggesting a crucial role of motor imagery in normal movement initiationand execution. Better understanding of motor imagery mechanisms inphysiologic and pathologic conditions is likely to lead to more eectiverehabilitative approaches.

Plasticity of emotional and cognitive functions

Integrated, noninvasive neuroimaging methods are useful in studyingcomplex cortical functions, such as emotions and cognition. Emotions playa fundamental role in survival and adaptation to the environment, and theyprovide much of the substrate for decision making, learning, and memory.They are fundamental for personality shaping during development andrecovery after a lesion [140]. Plasticity is likely to play an important role in theneural circuitry underlying emotions and emotional learning [141]. Plasticityphenomena most likely are occurring in response to early environmentalfactors, and they may be crucial in the expression of individual dierencesand risk for psychiatric disease [142]. In addition, psychiatric pharmacologicand nonpharmacologic interventions are likely to exert their eects largely byinducing plastic changes in the brain [143]. Better knowledge of theseprocesses could enhance the eectiveness of psychiatric treatments.

The amygdala, the hippocampus, and the prefrontal cortexthreestructures involved in emotional learning and personality and aectivestylesall are sites of neural plasticity (for a review, see Davidson et al [140]).The amygdala seems to be involved in learning especially for negativeemotionally charged stimuli. The amygdala is activated on fMRI and PETstudies in response to facial expressions of fear [144,145], aversive olfactoryand gustatory stimuli [146148], and unpleasant pictures [149] and in theearly phases of aversive conditioning [150152]. In general, the amygdalaseems to be fundamental for the development of conditioned fear [153] andthe autonomic responses elicited by fearful stimuli. Activation responses ofthe amygdala show a rapid habituation, and the long-term learning eectsmost likely are related to the connections between amygdala and prefrontalcortex [144]. Its main role might be expressed in the phase of associationbetween a stimulus and the threat it might pose and in the expression of fearin relationship to learned cues [140]. Abnormal activation of the amygdala inthe wrong context (eg, in response to neutral stimuli) is likely to subtendsome of the abnormal responses seen in phobic and anxiety-disorderedpatients. PET and fMRI studies of these patients have shown abnormalactivation in response to their specic anxiety-provoking stimuli comparedwith controls [154,155].

Animal studies have shown that early manipulations of the rearing

278 P.M. Rossini, G. Dal Forno / Phys Med Rehabil Clin N Am 15 (2004) 263306environment in rats produces a cascade of biologic changes that shape laterlife behavior and response to stress [142,156,157]. Changes in the receptordensity for benzodiazepines, corticotropin-releasing hormone, and gluco-

corticoids have been seen in the amygdala, hippocampus, and prefrontalcortex of developing animals, depending on the rearing behaviors (eg,licking, grooming) of the mothers [156]. Changes in corticotropin-releasinghormone and glucocorticoid receptors, in particular, are likely to modifysubstantially the response to stress-induced glucocorticoids. High circulat-ing glucocorticoid levels have been shown to accelerate neuronal loss in thehippocampus and to produce cognitive and aective impairment [158]. Inhumans and animals, neurogenesis and synaptic plastic changes are knownto occur in the dentate gyrus of the hippocampus even in adult life [159,160].Long-standing depression has been seen in association with measurablehippocampal atrophy on MRI, and these anatomic changes might explain inpart the cognitive abnormalities, particularly memory decits, seen inelderly depressed individuals [161164].

Hippocampal-related memory functions have been studied mostly usingPET and fMRI because of the deep location of this anatomic structure [165].Numerous studies exist on frontal cortexrelated memory and recalldomains. Repetitive TMS (rTMS) is particularly helpful in this eld for itsability to disrupt normal function when delivered over specic cortical areas.In animal cellular experiments, repetitive stimulation of the hippocampuswas shown to induce plastic synaptic changes and induce LTP at rapid ratesand LTD at low rates of stimulation [166]. Whether the same synapticchanges also happen when the human brain is stimulated in vivo with rTMSis not certain, although it has been shown that high-frequency rates ofstimulation result mainly in excitatory changes, whereas slow, low rates havea net inhibitory eect [167,168]. Disruption of free recall, working memory,and implicit learning has been shown with frontal cortex stimulation[169,170]. The right and left frontal lobes seem to display an asymmetricspecicity in encoding and retrieval mechanisms, as also suggested by PETneuroimaging studies of episodic memory [171177]. In a study on normaladults, the role of the right and left prefrontal cortex in long-term memorymechanisms was investigated by rTMS [178]. Disruption of the right or theleft dorsolateral prefrontal function was induced by high-frequency rTMS,while the subjects were involved in either the learning/encoding or retrievingphase of a visual learning task, whose stimuli consisted of pictures ofcomplex scenes. Left prefrontal rTMS resulted in impairment limited mostlyto the encoding portion of the task, whereas right stimulation was associatedwith disruption during the retrieval phases of the experiment. These ndingsnot only conrm the role of prefrontal cortex in long-term memoryfunctions, but also support the notion that an asymmetric functionalspecialization exists between the two hemispheres unrelated to language.

Of all cognitive functions, language and its plasticity have been studied

279P.M. Rossini, G. Dal Forno / Phys Med Rehabil Clin N Am 15 (2004) 263306extensively with functional neuroimaging methods, especially in patientsaected by various types of postlesional aphasias. Compensatory plasticityof residual tissues in language areas or extension of language function toareas previously not involved in language processing must exist to explain the

nearly universal improvement seen in postlesional aphasias, despitepersistence of the focal neuronal damage [179,180]. Evidence from morerecent studies shows that areas most likely to take over in aphasia recoveryare either the undamaged portions of the dominant hemisphere or thehomologous areas in the nondominant hemisphere [181]. In a PET studyperformed in the acute poststroke stage on patients with Wernickes aphasia,activation of the right temporoparietal region in left-dominant individualswas predictive of a favorable recovery of auditory comprehension [182].Similarly a PET study on recovered Wernickes aphasia patients performinga verb generation task showed activation of right superior temporal gyrus,inferior prefrontal and dorsolateral prefrontal cortex, and increasedactivation of residual left prefrontal cortical areas [183]. Better understand-ing of mechanisms possibly improving on naturally occurring restorativephenomena is needed to help rehabilitative eorts during postlesionallanguage retraining. Language trainingrelated improvement in comprehen-sion has been seen to correlate with measurable PET activation changes. Ina PET study of four recoveringWernickes aphasia patients, intense languagecomprehension training and relative task improvement resulted in measur-able short-term increase in blood ow in the left precuneus and right superiortemporal gyrus in 12 consecutive PET scans [184]. In particular, activation ofthe left precuneus, a structure normally involved in working memory andword retrieval [185], suggests the development of compensatory strategies,possibly by using structures not strictly viewed as language areas. Thesendings are important in light of the fact that it is not clear whethertreatment focused on a particular aspect of language could result inimprovement extending to other aspects of the language network [186].Therapy directed toward language recovery ideally should help a patientgeneralize his or her achievements to all or many aspects of language. Somestudies have shown that treatments focused on a particular aspect oflanguage can result in recruitment of other aspects of the language network[187189]. Certain compensatory strategies used to improve communicationthat bypass the language network may result in less than optimalimprovement and plastic neural network rearrangement [186].

In addition to blood owrelated functional neuroimaging techniques,rTMS is a valuable tool for studying plasticity in language because ofits ability to induce temporary enhancement or dysfunction of specicsubsystems. Cappa et al [190] used high-frequency rTMS in a verbal taskdesigned to evaluate the cortical areas involved in action naming (ie, theproduction of verbs, as opposed to nouns). The investigators found that onlythe stimulation of the left dorsolateral prefrontal cortex, as opposed to theright homologous regions or sham stimulation, induced facilitation of actionnaming, represented by shorter latency in the correct verbal reaction time.

280 P.M. Rossini, G. Dal Forno / Phys Med Rehabil Clin N Am 15 (2004) 263306Noninvasive functional neuroimaging techniques and their integratedapplication hold a great promise for better understanding of human cognitiveand mental/emotional processes. In addition, these techniques can help in

the development of rational rehabilitative and therapeutic approaches andin the monitoring of the results.

Peripheral and central lesions aecting sensorimotor area function

Merzenich et al [6,7] showed for the rst time that mechanisms similar tothose operating during early CNS development persist throughout life in theprojection systems of primates. Transection of the median nerve in adultmonkeys resulted, after several months, in new, expanded representationsof the skin elds in cortical areas 3b and 1, surrounding the arearepresenting the denervated skin district. No evidence of prior representa-tion of these skin surfaces was apparent before nerve transection. In thesame study, newly enlarged representations of the dorsal surfaces of digits 1and 2 (innervated by the radial nerve) and of the hypothenar eminence(innervated by the ulnar nerve) were shown. These expanded representa-tions appeared at times to move entirely into the former median nerverepresentational zone [7]. Similarly, 2 to 8 months after surgical amputationof digits in adult monkeys, the representations of adjacent intact digits andpalmar surfaces expanded topographically to occupy most or all of thedeaerented cortical territories, along with an increase in the corticalmagnication factor. Skin surfaces were represented in a correspondinglyner grain, implying that receptive eld overlap and separation in distanceacross the cortex were maintained dynamically. These changes in receptiveeld size had a spatial distribution and a time course similar to the changesin sensory acuity found on the stumps of human amputees [191].

After upper extremity amputation in humans, MEG studies of equivalentcurrent dipole (ECD) locations revealed that tactile lip stimulation evokedresponses in the cortical region normally corresponding to the amputatedhand, together with a signicant increase of the representation area of thengers of the intact hand. This change presumably resulted from a relativeincrease in sensory input to this area because of greater use of anddependence on the intact hand [192]. Reorganization of the motor systemcan occur 2 years after amputation. With TMS, a larger representation ofthe biceps muscle was found in seven of eight subjects with forearmamputation, a pattern less evident for muscles more proximal to theamputation, as measured by MEPs on the deltoid muscle [193]. In amputeesat the elbow, MEG revealed a strong correlation between amount of corticalinvasion by lip representation in the deaerented cortex and amount ofevoked pain sensation to the phantom limb [194]. Repeated examinations ofphantom limb patients showed a complete change of referred sensationtopography, implying a mislocalization phenomenon. Although the extent

281P.M. Rossini, G. Dal Forno / Phys Med Rehabil Clin N Am 15 (2004) 263306of the reorganization is a stable phenomenon, overall, these results suggestthat concomitant changes in sensory processing patterns are not stable.Alterations of sensory processing may be mediated by extensive, inter-connected neural networks with uctuating synaptic strengths [195].

PET measurements of regional cerebral blood ow in traumatic amputeeswith prominent phantom limb symptoms showed a ow increase of signif-icant magnitude in a wide area of contralateral MI/SI cortex, correspond-ing to the partially deaerented portions. These phenomena and theirneurophysiologic correlates provide examples of maladaptive plasticity. Incongenital amputees, without phantom limb symptoms, blood ow increasealso was present over a wide area in the partially deaerented MI/SI cortex,but not signicantly dierent in magnitude from the normally innervatedMI/SI cortex. TMS studies of traumatic amputees showed that the abnor-mal blood ow increase in the partially deaerented MI cortex is associatedwith increased corticospinal excitability [196].

Sensory cortical deaerentation aects cortical sensorimotor maps notonly when input is lacking for long periods, but also swiftly. Motor cortexTMS experiments showed that short-term deaerentation (mostly obtainedby peripheral ischemic nerve block) induced rapid increments of MEPamplitudes in the muscles proximal to the nerve block site, without anymodication in motor thresholds [197200]. These rapid plastic changes seemto be due to removal of GABAergic cortical inhibition and to sodiumchannel and calcium channel synaptic ecency shifts. The prolongedpersistence of the changes beyond the actual deaerentation period suggeststhat long-lasting plastic mechanisms are involved [200]. Rapid changes,induced by deaerentation with anesthetic block at the wrist of the medianand radial nerves, were shown in the cortical motor maps in normalvolunteers. The rst dorsal interosseus cortical representation was reducedsignicantly, despite maintaining its usual proprioceptive feedback andstrength via the ulnar nerve. The cortical representation of the abductordigiti minimi (serving as control condition because outside the anesthe-tized hand area) was unchanged, even if showing a tendency to enlarge in thetopographic maps. No amplitude changes of the MEPs of abductor digitiminimi and rst dorsal interosseus muscles were observed during peripheralstimulations, whereas changes in F-wave responses were detected in bothmuscles. No signicant topographic changes were found for the wrist exormuscles. These short-term rearrangements of brain motor maps seem tooccur on the basis of neural plasticity mechanisms secondary to the loss ofthe tonic cutaneous inputs on cortical and spinal motoneurons [201].

Spinal cord lesions and motoneuron diseases

In patients with traumatic lower cervical cord sections, focal magnetic coilstimulation elicited MEPs for most caudal muscles spared by the lesion

282 P.M. Rossini, G. Dal Forno / Phys Med Rehabil Clin N Am 15 (2004) 263306(biceps and deltoid) from wider scalp areas than in normal subjects. Latencyof biceps and deltoid MEPs were inversely related to amplitude. In patientswith traumatic quadriplegia, reorganization of the motor cortical projectionsystem can be shown by the fact that stimulation of areas normally eliciting

nger movements activate instead muscles just above the level of spinal cordlesion [202].

In patients with history of paralytic poliomyelitis and unilateral proximalinvolvement of the upper extremity, long-term rearrangement of the motorcortices was shown by signicant, asymmetric increase in excitable areas andamplitudes of theMEPs obtained withTMS of the contralateral motor cortexcompared with the ipsilateral motor cortex. This eect was evident only atthe level of the aected muscle representation (deltoid) and not in unaectedmotor distal muscles (abductor pollicis brevis) of the aected side [203].

Monohemispheric lesions and stroke

Numerous factors can inuence degree and mode of reorganization aftera stroke, such as site and extent of the lesion, diaschisis, and prestrokeorganization of the motor areas, in particular the amount of ipsilateraluncrossed corticospinal bers [204]. When damage to a functional systemis partial, a within-system recovery is possible, whereas after completedestruction, substitution by functionally related systems remains the onlyalternative [205]. Experimental and clinical studies have shown thatapproximately one fth of the pyramidal bers are sucient to ensurerestitution of fractionated hand nger movement. In this case, this within-pyramidal system reorganization is probably one of the best mechanisms forfunctional recovery of the motor control of hand and upper limb (Fig. 6)[206216].

Early PET and fMRI studies documented abnormal activation patternsduring movement of a paretic hand even after full clinical recovery [46,217220]. Enhanced bilateral activation of motor pathways has been reportedafter striatocapsular infarctions, together with recruitment of sensory andsecondary motor structures normally not involved in the task and dis-placement of primary motor peak activation toward the face area [220222].Because of their considerable logistic complexity, only a few longitudinalstudies have been published to date [221,223227].

In patients with cortical infarcts, a pattern of overactivation of bilateralnoninfarcted motor-related and nonmotor-related areas similar to that seenin striatocapsular strokes has been reported [217,218,228]. Additionalndings have been a strong peri-infarct activation [218] and ipsilesionalpremotor cortex activation [228]. The posterior shift and inferior extension ofMI activation observed in cortical and subcortical strokes might representunmasking/disinhibition of synaptic connections within the corticospinaltract system (Fig. 7).

Functional reorganization of the motor output after a hemispheric stroke

283P.M. Rossini, G. Dal Forno / Phys Med Rehabil Clin N Am 15 (2004) 263306has been observed consistently also via TMS and MEG [214,215,229,230].In stroke patients, MEPs recorded in a relaxed muscle state often are absentwith TMS of the aected hemisphere [231]. Other abnormalities includeincreased excitability threshold, delayed latency of the responses, and an

excessive asymmetry of the hand muscle motor maps between the aectedand the unaected hemispheres (AH and UH). These asymmetries are dueto map migration, usually on the mediolateral axis. Often an ante-roposterior shift of several millimeters of the CoG also has been reported.This shift of the CoG is particularly frequent and stable over time in patientsin the chronic stages of stroke, whereas it can be seen as a transient dynamicphenomenon in the rst few months after a stroke [207,214,215,229].

Both cerebral hemispheres, ipsilateral and contralateral to a stroke, mayplay an important role in functional recovery in the acute and subacutestages [46,232,233]. The remote eects of the lesioned area, or diaschisis,suggest that acute neuronal failure in the ischemic area induces modulatoryeects on cortical excitability of unaected districts of the same hemisphere,via corticocortical connections, and of the contralateral UH via trans-callosal bers [234,235].

Transient hyperexcitability of the UH contralateral to a neocortical

Fig. 7. The anatomic and functional correlates of the hand sensorimotor areas in a stroke patient

with a malacic lesion in the left frontoparietotemporal cortex. Transcranial magnetic stimulation

(TMS) mapping, functional MRI (fMRI), and magnetoencephalography (MEG) are integrated

as a method of functional imaging. Asymmetric enlargement and posterior shift of the

sensorimotor areas localized in the aected hemisphere are seen with all three techniques. (Data

from Rossini PM, Caltagirone C, Castriota-Scandberg A, et al. Hand motor cortical area

reorganization in stroke: a study with fMRI,MEG and TMSmaps. Neuroreport 1998;9:21416.)

284 P.M. Rossini, G. Dal Forno / Phys Med Rehabil Clin N Am 15 (2004) 263306infarction has been documented in animal models [236,237]. The relation-ship between clinical recovery and early cortical excitability to TMS aftera stroke has been studied as a possible prognostic tool [231,238240]. MEPsfrom upper limb muscles always have been obtained in patients who recover

nger control fully [231,241]. Less is known about excitability in subacuteand chronic stages (weeks to months) and correlation with clinical outcome.

The role of other nonpyramidal motor cortex eerents, such as thebilaterally organized premotor/reticulospinal projections, is still unclear.Large-scale reorganization of MI beyond the boundaries of the physiologiccorticocortical connections probably requires long-term repetitive engage-ment [45]. Consistent overrecruitment of motor and nonmotor cortical areashas been documented in AH and UH, with displacement of the MI peakactivation, in three fMRI studies [217,221,228] despite dierent experimentaldesigns. All the studies documented dynamic changes in activation degree orpattern as recovery proceeded. A common observation was that repeatedperformance of a task over time was associated with bihemisphericdampening of the hyperactivation. fMRI studies have shown consistenthyperactivation of peri-infarct cortex.

In humans, as in experimental animals, there is growing evidencesupporting a positive inuence of brain monoamine concentrations on rateand degree of recovery from cortical lesions [242]. Particularly if coupledwith physical therapy, amphetamine, levodopa, and uoxetine seem toenhance recovery [243,244], possibly via enhanced serotoninergic trans-mission, at least in the case of uoxetine. In an fMRI study, after a singledose of uoxetine, patients with subcortical strokes showed better motorperformance and activation increase in the ipsilesional sensorimotorprimary cortex when moving their paretic hand [245]. The degree ofactivation redistribution toward the MI output area was correlated linearlywith enhanced motor performance.

TMS studies performed in the acute (472 hours), subacute (1 week), andchronic stages (6 months) after a stroke showed increased excitabilitythresholds and prolonged MEP latencies (when present) and centralconduction time (CCT) acutely in the AH, whereas responses from theUH remained normal in each recording session [246,247]. Interhemisphericdierences between AH and UH were signicant for all the parametersstudied. The responses from the AH changed signicantly during follow-up,with most parameters showing a partial recovery: Latencies of MEPsdecreased and amplitudes increased, although they still were altered at 6months, but the clinical improvement still was ongoing. Interhemisphericdierences between AH and UH were the most altered neurophysiologicparameters, particularly for excitability thresholds. MEP amplitudes acutelyhad a positive correlation with clinical picture and with long-term outcome:The higher the MEP amplitude, the better the outcome. No correlation wasfound between side (right versus left) or type of lesion (cortical versussubcortical) and neurophysiologic parameters.

Hemispheric motor output also was studied at two dierent times after

285P.M. Rossini, G. Dal Forno / Phys Med Rehabil Clin N Am 15 (2004) 263306a stroke (T1 8 weeks; T2 1618 weeks). Excitability thresholds weresignicantly higher and the MEP amplitudes smaller in the AH despitea stronger TMS employed. The area of cortical output to the target muscle

was asymmetrically restricted compared with controls [248] and with thepatients UH. The percentage of altered parameters was signicantly higherin T1 and in patients with subcortical lesions, possibly as a result of the largenumber of densely packed bers aected by a subcortical lesion and a lessecient short-term plastic reorganization. Anomalous hot-spot sites wereobserved more frequently in the cortical group. In T2, the parameters ofsubcortical patients were improved to the same level of the cortical group.A signicant enlargement of the AH hand motor cortical area was found in67% of cases in T2 compared with T1, together with clinical improvement,as measured by the Canadian Neurological Scale, its Hand Motor subscore,and the Barthel Index. Patients showing enlargement of the hand area alsoshowed a correlated improvement of the hand score. Type of lesion, whethercortical or subcortical, did not inuence the clinical outcome (Fig. 8)[230,248]. When followed up in poststroke epochs, various MEP parameters

Fig. 8. Poststroke reorganization of brain motor output to the hand. Focal transcranial

magnetic stimulation (TMS) study of functional properties of hand motor areas 2 to 4 months

after a monohemispheric stroke. The two top gures show the TMS maps of aected and

unaected hemispheres, recorded at the beginning (T1) and after 8 to 10 weeks (T2) of

rehabilitation. In stroke patients, the aected hemisphere is less excitable than normal at T1

(smaller black dots), whereas at T2, the hand motor area on the aected hemisphere is enlarged,

corresponding to clinical improvement. The bottom gure shows TMS maps of healthy

286 P.M. Rossini, G. Dal Forno / Phys Med Rehabil Clin N Am 15 (2004) 263306controls. (From Cicinelli P, Traversa R, Oliveri M, Plamieri MG, Filippi MM, Rossini PM.

Neuropsychological marker or recovery of function after stroke. In: Reisin RC, Nuwer MR,

Hallett M, Medina C, editors. Advances in clinical neurophysiology. New York: Elsevier; 2002.

p. 23647; with permission.)

improved; the steepest parts of the recovery slopes were concentrated in the40 to 80 days after stroke [249].

The role of ipsilateral corticospinal connections is controversial, particu-larly their contribution to motor recovery after a unilateral lesion. In severalstudies, ipsilateral MEPs (iMEPs) could be elicited by TMS over the UH ofstroke patients [241,250254] or over the AH [254256]. The iMEPs usuallywere obtained stimulating anteriorly and medially to the primary motorcortex, suggesting activation of corticospinal pathways from premotor areas.Outcomewas variable; some investigators found a correlation between iMEPsand motor recovery [250,251,254], whereas others found no correlation[241,252,253]. The iMEPs described by Caramia et al [250] and Trompettoet al [254] seemed to dier: In Caramias patients, the iMEPS had lowexcitability thresholds and large amplitudes, while in Trompettos patients,only small iMEPs could be elicited with high stimulation intensities. Alagonaet al [255] found an association between iMEPs produced by stimulation of theAH and bimanual dexterity 6 months after stroke, suggesting that theirexistence reects hyperexcitability of AH premotor areas.

The study of intracortical inhibition (ICI) and facilitation [257] early aftera stroke sheds light on some plasticity mechanisms. A TMS study of largemiddle cerebral artery infarctions, with single and paired pulse TMS of theUH, showed a decrease of ICI 2 weeks after the stroke [258]. This ndingcorresponds to results obtained in animal studies [236,237,259]. In rats, theICI was associated with GABA (A) receptor down-regulation andglutamatergic activity enhancement [260,261]. Because ICI changes aresupposed to be modulated by GABAergic activity [262,263], the decreasedICI in the UH of patients with large infarctions may reect GABA activitydown-regulation. This eect could be due either to damage of transcallosalbers and loss of the physiologic interhemispheric inhibitory modulation[264267] or to enhanced use of the unaected arm in daily activities becauseICI is modied in a task-dependent and use-dependent manner [268]. Adisinhibition in the UH also is supported by the nding of enlarged MEPamplitudes when stimulating this hemisphere after a stroke [215,248,254].

A loss of ICI was observed in the AH in stroke patients with either smallmotor decits at onset or rapid spontaneous motor recovery, whereas ICI inthe UH was not dierent from age-matched controls [91,92,269,270]. Thisresult corresponds to animal studies describing loss of GABAergic inhibitionimmediately surrounding a cortical lesion [271]. On this basis, it has beensuggested that the rapid clinical improvement of some patients might beinduced by motor cortical disinhibition, secondary to the remote eects ofstructurally intact areas (diaschisis) or to the preexisting organization of themotor areas (ie, the amount of ipsilateral uncrossed corticospinal bers) [270].

Reorganization of hand and nger somatosensory areas has been

287P.M. Rossini, G. Dal Forno / Phys Med Rehabil Clin N Am 15 (2004) 263306investigated via MEG methods. Recordings of somatosensory evoked eldsover the contralateral parietal regions were performed 9 or more weeks aftera rst-ever monohemispheric stroke via separate electric stimulation of

median nerve, thumb, and little nger of both hands. Only the initial part ofthe somatosensory evoked elds, which contains the most stable, repeatable,and attention-independent waves (N20m, P30m), was analyzed. The ECDcharacteristics (spatial coordinates and strength) were calculated at 1-msecintervals in the 15- to 50-msec poststimulus epoch. The hand extensionarea also was calculated as the distance in millimeters between the centers ofthe ECDs activated by fth and rst nger stimulation. A MEG/brain MRIcommon reference system was dened on the basis of anatomic landmarks.All the somatosensory evoked eld parameters, including interhemisphericdierences, had been tested previously in a group of healthy subjects[213,272,273] and the abnormality threshold dened. Excessive interhemi-spheric signal strength asymmetry was found for at least one stimulateddistrict in 63% of patients with recordable somatosensory evoked elds fromthe AH, and nearly 80% of ECD pairs from homologous districts exceededthe normative range. When latency delays were accompanied by ECD spatialdisplacements, these always aected the N20 component generator sources.With the integration of MEG data with MRI anatomic information, it wasshown that all identied ECDs were localized outside the area of theanatomic lesion. The classic homunculus somatotopy, seen in the healthyhemisphere, was maintained even for the migrated sensory hand areas(thumb more lateral, little nger more medial, and median nerve in between).A signicant enlargement of the hand area was seen in the AH, with a medialshift of the little nger and a tendency of both nger representations to shiftanteriorly. Abnormal parameters were encountered infrequently in UH.Greater interhemispheric asymmetry, reective of cortical reorganization,correlated with worse clinical recovery, in agreement with longitudinalfunctional imaging studies of poststroke aphasia showing that betterrecovery was associated with return of left hemisphere perilesional activity,whereas persistent activation of homologous areas in the nondominanthemisphere probably represented maladaptive plasticity (Fig. 9) [181,217].

One way to examine specically intervention-induced plastic reorganiza-tion is to study patients in the chronic stages of their illness, a phase whenthe probability of spontaneous recovery is negligible [267269]. Anotherpossibility is to study the short-term eects of interventions (eg, within1 day), where changes observed after a single therapeutic session are likelydue to the intervention itself rather than to spontaneous improvements.Some studies have looked at the intervention-induced eects on motorreorganization of constraint-induced movement therapy (CIMT) [267,268,274276]. With TMS, repeated baseline measurements have shownhighly reproducible results in strokes that occurred more than 6 monthsprior. Before CIMT, the patients had higher motor thresholds and smallermotor output maps in the AH. After therapy, motor output maps in the AH

288 P.M. Rossini, G. Dal Forno / Phys Med Rehabil Clin N Am 15 (2004) 263306had increased by approximately 40%, whereas motor output maps in theUH had decreased, although not signicantly. The AH cortical map hadbecome larger than that in the UH. Both changes presumably reect use-

dependent mechanisms (ie, increased use of the paretic hand during trainingand decreased use of the nonaected hand). Neurophysiologic changes wereparalleled by a large clinical improvement of motor function. Motorthresholds had remained identical after CIMT, and because they aredetermined in the center of the cortical representation area, it was concluded

Fig. 9. Functional information from magnetoencephalography (MEG) somatosensory evoked

elds (SEFs), integrated with MRI anatomic data in the study of reorganization of sensory

hand areas 12 months after monohemispheric stroke, showing increased interhemispheric

asymmetry. (Left) (a) MRI scan of brain showing a subcortical stroke. (b) MEG/MRI

integration of the equivalent current dipoles (ECDs) in the unaected hemisphere (UH). (c)

MEG/MRI integration of the ECDs in the aected hemisphere (AH). (d ) Diagram of ECDs in

UH and AH [212]. (Right) MRI of a left frontotemporoparietal ischemic lesion (top). MEG/

MRI integration shows asymmetric location of ECDs activated by contralateral fth nger

stimulation in the AH (middle) with respect to the UH (bottom) [277]. (Data from Rossini PM,

Tecchio F, Pizzella V, et al. On the reorganization of sensory hand areas after mono-

hemispheric lesion: a functional (MEG)/anatomical (MRI) integrative study. Brain Res

1998;782:15366. Rossini PM, Tecchio F, Pizzella V, et al. Interhemispheric dierences of

sensory hand areas after monohemispheric stroke: MEG/MRI integrative study. Neuroimage

2001;14:47485.)

289P.M. Rossini, G. Dal Forno / Phys Med Rehabil Clin N Am 15 (2004) 263306that enlargements of the motor output maps were due to excitabilityincrease at the borders of the representation area, possibly through GABA-dependent modulation of horizontal intracortical inhibitory circuits [13].

After CIMT, the CoG of the motor output maps had shifted signicantlyin the AH mainly in the mediolateral axis, presumably indicatingrecruitment of additional adjacent brain areas. Similar results have beenreported in monkey studies [2,278]. In the subsequent 6 months after CIMT,the motor output map sizes equalized (\ area in AH, > area in UH),whereas motor performance remained unchanged. These changes wereinterpreted as indicators of improved connectivity between neuronalpopulations, allowing a reduction of cortical excitability while maintainingperformance [267]. In contrast, an MEG study found a large shift ofmovement-related cortical potentials in the UH 3 months after CIMT [274].CIMT and manipulation of monoamine neuromodulators have been shownin randomized controlled trials to enhance motor recovery via stimulationof use-dependent connections [267,268,275,276]. The facilitatory eects ofdierent interventions studied with TMS showed that contraction of thetarget muscle exceeded all other facilitatory techniques, such as preinnerva-tion of more proximal or contralateral homologous muscles or passivecutaneous stimulation [279]. These results conrm that increased orrepetitive use has facilitatory eects and improves motor performance [280].

fMRI studies reported that specic rehabilitative procedures, eitherpassive or active, are able to induce signicant changes in brain activation. Inpatients with recent subcortical strokes, 3 weeks of intensive rehabilita-tioncompared with standard stimulationresulted in enhanced activationof ipsilesional sensorimotor primary cortices [281]. Similarly, in a group ofpatients with cortical or subcortical chronic stroke, predominant activationof contralesional MI cortex during active motor tasks before training evolvedin a preferential ipsilesional MI activation with intense training of the pareticarm and resulted in signicant clinical improvement [282].

Integrated methods of functional brain imaging have been employed onlyin a few studies. In a paradigmatic case, fMRI, TMS, and MEG all agreedin showing asymmetric enlargement and posterior shift of the sensorimotorareas of the AH [230]. Interhemispheric dierences also have been evaluatedwith fMRI. Data were analyzed using a laterality index, ranging from +1 to1, as an indicator of shift of the motor network activation toward the UH,and spatial coordinates to measure the displacement of peak activation ofMI [218,223,225,281]. In clinically stabilized chronic strokes, laterality indexis distributed more broadly compared with controls, owing to prominentcontralesional activation, whereas in controls there is preservation of thenormal interhemispheric balance. One study [223] observed a positivecorrelation between laterality index changes over time and concomitantnger motion recovery, documenting that the greater the activation shifttoward the UH (greater laterality index value), the worse the recovery.

As a general comment in regard to fMRI and PET studies in stroke, one

290 P.M. Rossini, G. Dal Forno / Phys Med Rehabil Clin N Am 15 (2004) 263306might argue that motor paradigms using nger tapping usually have beenemployed. In the authors opinion, this practice introduces potential biaslinked to the dierent strategies of movement programming, execution, and

feedback from the paretic hand with respect to the unaected one. For thisreason, objective types of activation methods (eg, electric shocks) should beapplied when comparing responsiveness of the two hemispheres andhemispheric asymmetries of the evaluated parameters. Interhemisphericdiierences also were of interest in MEG studies. For the hand area, inparticular, parameters were abnormal in 20% and 13% of subcortical andcortical lesions due to enlargement of the hand area size (39 7 mm insubcortical strokes and 27 mm in cortical lesions versus 16 5 mm in healthycontrols). The area enlargement did not correlate signicantly with clinicaloutcome.

All these ndings from PET, fMRI, TMS, and MEG studies suggest thatreorganization of the motor output from a lesioned hemisphere still is takingplace several months after a stroke. The time course and degree of motorrecovery in humans could depend largely on the degree of damage to thepreviously described distributed motor network because dierent motorareas operate in a parallel rather than in a hierarchical fashion, and paralleldescending pathways might be able to compensate functionally for eachother [41]. The poststroke interval of some studies was long enough tosuggest that the observed modications were due to corticospinal tractreorganization, rather than recovery from perilesional edema and earlycortical hypoexcitability. Recovery of sensory decits also can playa signicant role because the modulation of the tonic sensory ow fromthe skin enveloping the target muscle signicantly aects the amount of itscortical representation [201,283].

Summary

The study of neural plasticity has expanded rapidly in the past decadesand has shown the remarkable ability of the developing, adult, and agingbrain to be shaped by environmental inputs in health and after a lesion[284]. Robust experimental evidence supports the hypothesis that neuronalaggregates adjacent to a lesion in the sensorimotor brain areas can take overprogressively the function previously played by the damaged neurons. Itdenitely is accepted that such a reorganization modies sensibly theinterhemispheric dierences in somatotopic organization of the sensorimo-tor cortices. This reorganization largely subtends clinical recovery of motorperformances and sensorimotor integration after a stroke.

Brain functional imaging studies show that recovery from hemiplegicstrokes is associated with a marked reorganization of the activation patternsof specic brain structures [46]. To regain hand motor control, the recoveryprocess tends over time to bring the bilateral motor network activation

291P.M. Rossini, G. Dal Forno / Phys Med Rehabil Clin N Am 15 (2004) 263306toward a more normal intensity/extent, while overrecruiting simultaneouslynew areas, perhaps to sustain this process. Considerable intersubjectvariability exists in activation/hyperactivation pattern changes over time.

Some patients display late-appearing dorsolateral prefrontal cortex activa-tion, suggesting the development of executive strategies to compensate forthe lost function. The AH in stroke often undergoes a signicant remodel-ing of sensory and motor hand somatotopy outside the normal areas, orenlargement of the hand representation. The UH also undergoes re-organization, although to a lesser degree. Although absolute values of theinvestigated parameters uctuate across subjects, secondary to individualanatomic variability, variation is minimal with regards to interhemisphericdierences, due to the fact that individual morphometric characters aremirrored in the two hemispheres. Excessive interhemispheric asymmetry ofthe sensorimotor hand areas seems to be the parameter with highestsensitivity in describing brain reorganization after a monohemispheric lesion,and mapping motor and somatosensory cortical areas through focal TMS,fMRI, PET, EEG, and MEG is useful in studying hand representationand interhemispheric asymmetries in normal and pathologic conditions[221,223,285].

TMS and MEG allow the detection of sensorimotor areas reshaping, asa result of either neuronal reorganization or recovery of the previouslydamaged neural network. These techniques have the advantage of hightemporal resolution but also have limitations. TMS provides only bidimen-sional scalp maps, whereas MEG, even if giving three-dimensional mappingof generator sources, does so by means of inverse procedures that rely on thechoice of a mathematical model of the head and the sources. Thesetechniques do not test movement execution and sensorimotor integration asused in everyday life. fMRI and PET may provide the ideal means tointegrate the ndings obtained with the other two techniques. Thismultitechnology combined approach is at present the best way to test thepresence and amount of plasticity phenomena underlying partial or totalrecovery of several functions, sensorimotor above all.

Dynamic patterns of recovery are emerging progressively from therelevant literature. Enhanced recruitment of the aected cortex, be it sparedperilesional tissue, as in the case of cortical stroke, or intact but deaerentedcortex, as in subcortical strokes, seems to be the rule, a mechanism especiallyimportant in early postinsult stages. The transfer over time of preferentialactivation toward contralesional cortices, as observed in some cases, seems,however, to reect a less ecient type of plastic reorganization, with someaspects of maladaptive plasticity. Reinforcing the use of the aected side cancause activation to increase again in the aected side with a correspondingenhancement of clinical function. Activation of the UH MI may representrecruitment of direct (uncrossed) corticospinal tracts and relate more tomirror movements, but it more likely reects activity redistribution withinpreexisting bilateral, large-scale motor networks. Finally, activation of areas

292 P.M. Rossini, G. Dal Forno / Phys Med Rehabil Clin N Am 15 (2004) 263306not normally engaged in the dysfunctional tasks, such as the dorsolateralprefrontal cortex or the superior parietal cortex in motor paralysis, mightreect the implication of compensatory cognitive strategies.

An integrated approach with technologies able to investigate functionalbrain imaging is of considerable value in providing information on theexcitability, extension, localization, and functional hierarchy of corticalbrain areas. Deepening knowledge of the mechanisms regulating the long-term recovery (even if partial), observed for most neurologic sequelae afterneural damage, might prompt newer and more ecacious therapeutic andrehabilitative strategies for neurologic diseases.

References

[1] Twitchell TE. The restoration of motor function following hemiplegia in man. Brain 1951;

74:44380.

[2] Nudo RJ, Milliken GW. Reorganization of movement representations in primary motor

cortex following focal ischemic infarcts in adult squirrel monkeys. J Neurophysiol 1996;

75:21449.

[3] Nudo RJ. Remodeling of cortical motor representations after stroke: implications for

recovery from brain damage. Mol Psychiatry 1997;2:18891.

[4] Taub E, Uswatte G, Elbert T. New treatments in neurorehabilitation founded on basic

research. Nat Rev Neurosci 2002;3:22836.

[5] Rossini PM. Brain redundancy: responsivity or plasticity? Ann Neurol 2001;48:1289.

[6] Merzenich MM, Kaas JH, Wall J, Nelson RJ, Sur M, Felleman D. Topographic

reorganization of somatosensory cortical areas 3b and 1 in adult monkeys following

restricted deaerentation. Neuroscience 1983;8:3355.

[7] Merzenich MM, Kaas JH, Wall JT, Sur M, Nelson RJ, Felleman DJ. Progression of

change following median nerve section in the cortical representation of the hand areas 3b

and 1 in adult owl and squirrel monkeys. Neuroscience 1983;10:63965.

[8] Xerri J, Merzenich MM, Peterson BE, Jenkins W. Plasticity of primary somatosensory

cortex paralleling sensorimotor skill recovery from stroke in adult monkeys. J Neuro-

physiol 1998;79:211948.

[9] Pascual-Leone A, Grafman J, Hallet M. Modulation of cortical motor output maps

during development of implicit and explicit knowledge. Science 1994;263:12879.

[10] Pascual-Leone A, Nguyet D, Cohen LG, Brasil-Neto JP, Cammarota A, Hallett M.

Modulation of muscle responses evoked by transcranial magnetic stimulation during the

acquisition of new ne motor skills. J Neurophysiol 1995;74:103745.

[11] Elbert T, Pantev C, Weinbruck C, Rockstroh B, Taub E. Increased cortical representation

of the ngers of the left hand in string players. Science 1995;270:3057.

[12] Hebb DO. The organization of behavior. New York: Wiley; 1949.

[13] Jacobs KM, Donoghue JP. Reshaping the cortical motor map by unmasking latent

intracortical connections. Science 1991;251:9447.

[14] Halter JA, Carp JS, Wolpaw JR. Operantly conditioned motoneuron plasticity: possible

role of sodium channels. J Neurophysiol 1995;73:86771.

[15] Sakamoto T, Porter LL, Asanuma H. Long-lasting potentiation of synaptic potentials in

the motor cortex produced by stimulation of the sensory cortex in the cat: a basis of motor

learning. Brain Res 1987;413:3604.

[16] Hess G, Jacobs KM, Donoghue JP. N-Methyl-D-aspartate receptor mediated compo-

nents of eld potentials evoked in horizontal pathways of rat motor cortex. Neuroscience

293P.M. Rossini, G. Dal Forno / Phys Med Rehabil Clin N Am 15 (2004) 2633061994;61:22535.

[17] Bliss TV, Lomo T. Long-lasting potentiation of synaptic transmission in the dentate area

of the anaesthetized rabbit following stimulation of the perforant path. J Physiol 1973;

232:33156.

[18] Buonomano DV, Merzenich MM. Net interaction between dierent forms of short-term

synaptic plasticity and slow-IPSPs in the hippocampus and auditory cortex.

J Neurophysiol 1998;80:176574.

[19] Asanuma H, Pavlides C. Neurobiological basis of motor learning in mammals.

Neuroreport 1997;8:IVI.

[20] Bindman LJ, Murphy KP, Pockett S. Postsynaptic control of the induction of long-term

changes in ecacy of transmission at neocortical synapses in slices of rat brain.

J Neurophysiol 1988;60:105365.

[21] Crair MC, Malenka RC. A critical period for long-term potentiation at thalamocortical

synapses. Nature 1995;375:3258.

[22] Rauschecker JP. Mechanisms of visual plasticity: Hebb synapses, NMDA receptors, and

beyond. Physiol Rev 1991;71:587615.

[23] Rauschecker JP. Developmental plasticity and memory. Behav Brain Res 1995;66:712.

[24] Kilgard MP, Merzenich MM. Cortical map reorganization enabled by nucleus basalis

activity. Science 1998;279:17148.

[25] Kirkwood A, Rozas C, Kirkwood J, Perez F, Bear MF. Modulation of long term synaptic

de