225 Neocortical mechanisms in motor learning Jerome N Sanes The ability to learn novel motor skills has fundamental structures have also appeared recently, but these are not importance for adaptive behavior. Neocortical mechanisms discussed in copious detail. support human motor skill learning, from simple practice to adaptation and arbitrary sensory–motor associations. Behavioral Any discussion of motor learning necessarily begins with and neural manifestations of motor learning evolve in time and defining its scope. Most commonly, motor learning entails involve multiple structures across the neocortex. Modifications acquiring novel movement patterns. Such novel patterns of neural properties, synchrony and synaptic efficacy are all can be the straightforward result of a simple action being related to the development and maintenance of motor skill. repeated and the involved muscles gradually developing a stereotyped pattern of agonist and antagonist activity [4]. Addresses Developing more complex skills involving multiple joints Department of Neuroscience, Brown Medical School, Box 1953, and across-limb coordination, for example when training Providence, RI 02912, USA for sport or artistic activities, entails the establishment of e-mail: [email protected] novel simultaneous and sequential action patterns. The category of motor learning also subsumes the acquisition Current Opinion in Neurobiology 2003, 13:225–231 of novel associations between environmental events and motor actions when, for example, learning to manipulate a This review comes from a themed issue on tool or responding for the first time to a traffic signal. The Cognitive neuroscience Edited by Brian Wandell and Anthony Movshon component actions — grasping, applying forces etc — and their sensory antecedents have already been acquired or 0959-4388/03/$ – see front matter experienced but not necessarily conjoined in a new con- ß 2003 Elsevier Science Ltd. All rights reserved. text. Regardless of whether the learning necessitates DOI 10.1016/S0959-4388(03)00046-1 adaptation or the formation of new sensory–motor rela- tionships, new patterns of neural activity or activation accompany these changes. Understanding the behavioral Abbreviations ERP event-related potential and neural concomitants of sequential actions has been a IFG inferior frontal gyrus fecund area of recent motor-learning research and will be IT inferior temporal cortex considered here. LTD long-term depression LTP long-term potentiation M1 primary motor cortex Movement practice PFC prefrontal cortex The simplest type of motor learning entails practice- PMA premotor area related performance changes. Numerous prior studies rTMS repetitive transcranial magnetic stimulation have indicated a neocortical role in these changes both in SMA supplementary motor area the short-term (minutes to hours) [5] and in the longer- SRTT serial reaction-time task term (hours to days or months and beyond) [5–7].M1 appears to figure prominently in the development of practice-related motor skills, as might be predicted from Introduction its propensity to exhibit rapid shifts in output represen- The proposition that the neocortex contributes to motor tation patterns [3]. Prior data indicated that simply plasticity and motor skill learning should provoke little repeating a particular finger movement was sufficient debate. That motor-related structures of the frontal lobe, to alter the M1 output for tens of minutes [8], although such as the primary motor cortex (M1), premotor area the specificity of the modified M1 output pattern was not (PMA) and supplementary motor area (SMA), have sig- established. More recent data have extended these find- nificant roles in the phenomena of motor learning might ings to indicate that M1 practice-related plasticity occurs have once engendered significant controversy. However, for rapid-ramp but not slow-ramp movements, suggest- the historical and recent record provide incontrovertible ing a functional specialization for M1 [9].Furthermore,it evidence that many neocortical regions, including the appears that M1 has a role in the early consolidation of motor-related areas, exhibit plasticity and are likely to practice-related changes in motor skills [10]. Disrup- contribute to motor-skill learning [1–3]. There have been tion of M1 processing by repetitive transcranial magnetic significant recent advances in detailing how the neocortex stimulation (rTMS) substantially reduced practice- contributes to motor-skill learning and these are reviewed related performance improvements. These effects did here with a focus on human physiology. Many important not seem to be related to M1 output, as this appeared findings relevant to motor learning involving subcortical normal upon testing with standard TMS evaluation, nor www.current-opinion.com Current Opinion in Neurobiology 2003, 13:225–231 226 Cognitive Neuroscience did they extend beyond the disruption of M1, as rTMS serial reaction-time task (SRTT), which has implicit delivered to the prefrontal cortex (PFC) or occipital and explicit components [18], although other equally cortex had no effect. valid tasks have been employed [19]. The SRRT typi- cally requires participants to respond discretely to each of Practice-related effects extend beyond the repetition of several visually presented cues. With repetition of simple movements and beyond changes in M1. Humans ordered sequences of stimulus-response pairs, decreases can perform rhythmic movements paced by a metronome in reaction time to the cue indicate implicit learning, in a syncopated or synchronized mode, although syncopa- whereas knowledge about the sequence order indicates tion performance degrades as the beat frequency explicit learning. increases. Syncopation ability improves with practice, which allows the task to be carried out at higher rhythmic Prior work using the SRTT has determined that wide- frequencies; this makes the transition from syncopation to spread neocortical networks become activated during the synchronized pacing a useful tool for studying how prac- implicit and explicit phases of the task. Subcortical tice might alter brain function. A greater ability to per- structures, including those in the basal ganglia and cer- form syncopated movements reduces the difference ebellum, also become activated during learning accom- between the magnetic field power recorded over the panying the SRTT. Despite the wealth of information a and b bands of the sensorimotor cortex during synco- about the neural substrates of the SRTT, numerous pated versus synchronized rhythm production [11]. questions remain unanswered. These uncertainties Using functional MRI, the same group assessed neocor- include whether the same areas mediate learning and tical network activation patterns in relation to improved expression of implicit and explicit knowledge, and ability to perform syncopated movements. Prior to prac- whether brain areas implicated in mediating SRTT have tice, syncopation activated a wider set of structures, a general or specific role in variants of the task; that is, including the SMA, inferior frontal gyrus (IFG), superior sequence learning that involves spatial, temporal or temporal gyrus and subcortical structures, than the syn- object attributes of the stimuli cueing movement or of chronization task [12]. This observation could relate to the movement itself. the increased complexity of the syncopation task, which hence requires greater attention to sensory–motor timing. Prior results have indicated that different brain networks Analogous to findings in other domains, for example become activated during the development of implicit and speech [13], practice on the syncopation task reduced explicit knowledge while the SRTT is being performed activation in the SMA and IFG while the task was being [20–22]. Implicit knowledge activated neocortical motor- carried out, resulting in the two tasks displaying more related areas, including M1 and SMA; this is consistent similar activation patterns. Practice-related changes in with the growth of M1 movement representation during neocortical activation also occur with targeted movement the development of implicit knowledge [23]. In contrast, tasks, where improvements in accuracy are observed. In a attainment of explicit knowledge yielded activation in recent event-related potential (ERP) study, the ampli- the PFC, PMA, and posterior parietal cortex, among other tude of frontal ERPs decreased whereas parietal ERPs areas. Implicit knowledge accrual for motor sequences increased with improvements in accuracy [14],afinding may extend beyond M1 and SMA, as recent data indicate analogous to a frontal-to-parietal shift in activation pat- activation in the ventral PMA, mid-cingulate, PFC, and tern with the acquisition of motor sequences by trial-and- posterior parietal cortex while implicit sequence knowl- error [15] or motor adaptation [16]. Taken together, it edge develops [24]. The involvement of parietal and appears that practice yields convergent neocortical pro- frontal networks during the rapid development of explicit cessing resembling that required for the simplest possible knowledge has been confirmed [25,26]. Evidence from action, a process that might be described as ‘finding the neuropsychology also suggests segregation of the neural lowest common denominator’ and that is coupled with a mechanisms for implicit and explicit motor sequence shift in activation from frontal to parietal regions. learning. Patients with damage to the left parietal lobe fail