Skip to main content

Thank you for visiting You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Emergence of reproducible spatiotemporal activity during motor learning


The motor cortex is capable of reliably driving complex movements1,2 yet exhibits considerable plasticity during motor learning3,4,5,6,7,8,9,10. These observations suggest that the fundamental relationship between motor cortex activity and movement may not be fixed but is instead shaped by learning; however, to what extent and how motor learning shapes this relationship are not fully understood. Here we addressed this issue by using in vivo two-photon calcium imaging11 to monitor the activity of the same population of hundreds of layer 2/3 neurons while mice learned a forelimb lever-press task over two weeks. Excitatory and inhibitory neurons were identified by transgenic labelling12,13. Inhibitory neuron activity was relatively stable and balanced local excitatory neuron activity on a movement-by-movement basis, whereas excitatory neuron activity showed higher dynamism during the initial phase of learning. The dynamics of excitatory neurons during the initial phase involved the expansion of the movement-related population which explored various activity patterns even during similar movements. This was followed by a refinement into a smaller population exhibiting reproducible spatiotemporal sequences of activity. This pattern of activity associated with the learned movement was unique to expert animals and not observed during similar movements made during the naive phase, and the relationship between neuronal activity and individual movements became more consistent with learning. These changes in population activity coincided with a transient increase in dendritic spine turnover in these neurons. Our results indicate that a novel and reproducible activity–movement relationship develops as a result of motor learning, and we speculate that synaptic plasticity within the motor cortex underlies the emergence of reproducible spatiotemporal activity patterns for learned movements. These results underscore the profound influence of learning on the way that the cortex produces movements.

This is a preview of subscription content

Access options

Buy article

Get time limited or full article access on ReadCube.


All prices are NET prices.

Figure 1: Lever-press task and chronic calcium imaging of excitatory and inhibitory populations.
Figure 2: Dynamics of spatiotemporal activity of excitatory neurons during learning.
Figure 3: Learning-related emergence of reproducible spatiotemporal activity.
Figure 4: Learning-related plasticity of dendritic spines.


  1. Graziano, M. S. A. The Intelligent Movement Machine 1st edn (Oxford, 2009)

    Book  Google Scholar 

  2. Harrison, T. C., Ayling, O. G. & Murphy, T. H. Distinct cortical circuit mechanisms for complex forelimb movement and motor map topography. Neuron 74, 397–409 (2012)

    CAS  Article  Google Scholar 

  3. Xu, T. et al. Rapid formation and selective stabilization of synapses for enduring motor memories. Nature 462, 915–919 (2009)

    ADS  CAS  Article  Google Scholar 

  4. Yang, G., Pan, F. & Gan, W. B. Stably maintained dendritic spines are associated with lifelong memories. Nature 462, 920–924 (2009)

    ADS  CAS  Article  Google Scholar 

  5. Nudo, R. J., Milliken, G. W., Jenkins, W. M. & Merzenich, M. M. Use-dependent alterations of movement representations in primary motor cortex of adult squirrel monkeys. J. Neurosci. 16, 785–807 (1996)

    CAS  Article  Google Scholar 

  6. Rioult-Pedotti, M. S., Friedman, D. & Donoghue, J. P. Learning-induced LTP in neocortex. Science 290, 533–536 (2000)

    ADS  CAS  Article  Google Scholar 

  7. Komiyama, T. et al. Learning-related fine-scale specificity imaged in motor cortex circuits of behaving mice. Nature 464, 1182–1186 (2010)

    ADS  CAS  Article  Google Scholar 

  8. Huber, D. et al. Multiple dynamic representations in the motor cortex during sensorimotor learning. Nature 484, 473–478 (2012)

    ADS  CAS  Article  Google Scholar 

  9. Sanes, J. N. & Donoghue, J. P. Plasticity and primary motor cortex. Annu. Rev. Neurosci. 23, 393–415 (2000)

    CAS  Article  Google Scholar 

  10. Rokni, U., Richardson, A. G., Bizzi, E. & Seung, H. S. Motor learning with unstable neural representations. Neuron 54, 653–666 (2007)

    CAS  Article  Google Scholar 

  11. Stosiek, C., Garaschuk, O., Holthoff, K. & Konnerth, A. In vivo two-photon calcium imaging of neuronal networks. Proc. Natl Acad. Sci. USA 100, 7319–7324 (2003)

    ADS  CAS  Article  Google Scholar 

  12. Taniguchi, H. et al. A resource of Cre driver lines for genetic targeting of GABAergic neurons in cerebral cortex. Neuron 71, 995–1013 (2011)

    CAS  Article  Google Scholar 

  13. Madisen, L. et al. A robust and high-throughput Cre reporting and characterization system for the whole mouse brain. Nature Neurosci. 13, 133–140 (2010)

    CAS  Article  Google Scholar 

  14. Isomura, Y., Harukuni, R., Takekawa, T., Aizawa, H. & Fukai, T. Microcircuitry coordination of cortical motor information in self-initiation of voluntary movements. Nature Neurosci. 12, 1586–1593 (2009)

    CAS  Article  Google Scholar 

  15. Hira, R. et al. Spatiotemporal dynamics of functional clusters of neurons in the mouse motor cortex during a voluntary movement. J. Neurosci. 33, 1377–1390 (2013)

    CAS  Article  Google Scholar 

  16. Kato, H. K., Chu, M. W., Isaacson, J. S. & Komiyama, T. Dynamic sensory representations in the olfactory bulb: modulation by wakefulness and experience. Neuron 76, 962–975 (2012)

    CAS  Article  Google Scholar 

  17. Weiler, N., Wood, L., Yu, J., Solla, S. A. & Shepherd, G. M. Top-down laminar organization of the excitatory network in motor cortex. Nature Neurosci. 11, 360–366 (2008)

    CAS  Article  Google Scholar 

  18. Kaneko, T., Cho, R., Li, Y., Nomura, S. & Mizuno, N. Predominant information transfer from layer III pyramidal neurons to corticospinal neurons. J. Comp. Neurol. 423, 52–65 (2000)

    CAS  Article  Google Scholar 

  19. Akerboom, J. et al. Optimization of a GCaMP calcium indicator for neural activity imaging. J. Neurosci. 32, 13819–13840 (2012)

    CAS  Article  Google Scholar 

  20. Isaacson, J. S. & Scanziani, M. How inhibition shapes cortical activity. Neuron 72, 231–243 (2011)

    CAS  Article  Google Scholar 

  21. Kwan, A. C. & Dan, Y. Dissection of cortical microcircuits by single-neuron stimulation in vivo. Curr. Biol. 22, 1459–1467 (2012)

    CAS  Article  Google Scholar 

  22. Kätzel, D., Zemelman, B. V., Buetfering, C., Wolfel, M. & Miesenbock, G. The columnar and laminar organization of inhibitory connections to neocortical excitatory cells. Nature Neurosci. 14, 100–107 (2011)

    Article  Google Scholar 

  23. Chestek, C. A. et al. Single-neuron stability during repeated reaching in macaque premotor cortex. J. Neurosci. 27, 10742–10750 (2007)

    CAS  Article  Google Scholar 

  24. Georgopoulos, A. P., Schwartz, A. B. & Kettner, R. E. Neuronal population coding of movement direction. Science 233, 1416–1419 (1986)

    ADS  CAS  Article  Google Scholar 

  25. Mandelblat-Cerf, Y., Paz, R. & Vaadia, E. Trial-to-trial variability of single cells in motor cortices is dynamically modified during visuomotor adaptation. J. Neurosci. 29, 15053–15062 (2009)

    CAS  Article  Google Scholar 

  26. Arduin, P. J., Fregnac, Y., Shulz, D. E. & Ego-Stengel, V. “Master” neurons induced by operant conditioning in rat motor cortex during a brain-machine interface task. J. Neurosci. 33, 8308–8320 (2013)

    CAS  Article  Google Scholar 

  27. Shenoy, K. V., Sahani, M. & Churchland, M. M. Cortical control of arm movements: a dynamical systems perspective. Annu. Rev. Neurosci. 36, 337–359 (2013)

    CAS  Article  Google Scholar 

  28. Long, M. A., Jin, D. Z. & Fee, M. S. Support for a synaptic chain model of neuronal sequence generation. Nature 468, 394–399 (2010)

    ADS  CAS  Article  Google Scholar 

  29. Krings, T. et al. Cortical activation patterns during complex motor tasks in piano players and control subjects. A functional magnetic resonance imaging study. Neurosci. Lett. 278, 189–193 (2000)

    ADS  CAS  Article  Google Scholar 

  30. Picard, N., Matsuzaka, Y. & Strick, P. L. Extended practice of a motor skill is associated with reduced metabolic activity in M1. Nature Neurosci. 16, 1340–1347 (2013)

    CAS  Article  Google Scholar 

  31. Holtmaat, A. et al. Long-term, high-resolution imaging in the mouse neocortex through a chronic cranial window. Nature Protocols 4, 1128–1144 (2009)

    CAS  Article  Google Scholar 

  32. Li, C. X. & Waters, R. S. Organization of the mouse motor cortex studied by retrograde tracing and intracortical microstimulation (ICMS) mapping. Can. J. Neurol. Sci. 18, 28–38 (1991)

    CAS  Article  Google Scholar 

  33. Pronichev, I. V. & Lenkov, D. N. Functional mapping of the motor cortex of the white mouse by a microstimulation method. Neurosci. Behav. Physiol. 28, 80–85 (1998)

    CAS  Article  Google Scholar 

  34. Ferezou, I. et al. Spatiotemporal dynamics of cortical sensorimotor integration in behaving mice. Neuron 56, 907–923 (2007)

    CAS  Article  Google Scholar 

  35. Ayling, O. G., Harrison, T. C., Boyd, J. D., Goroshkov, A. & Murphy, T. H. Automated light-based mapping of motor cortex by photoactivation of channelrhodopsin-2 transgenic mice. Nature Methods 6, 219–224 (2009)

    CAS  Article  Google Scholar 

  36. Tennant, K. A. et al. The organization of the forelimb representation of the C57BL/6 mouse motor cortex as defined by intracortical microstimulation and cytoarchitecture. Cereb. Cortex 21, 865–876 (2011)

    Article  Google Scholar 

  37. Olsen, S. R., Bortone, D. S., Adesnik, H. & Scanziani, M. Gain control by layer six in cortical circuits of vision. Nature 483, 47–52 (2012)

    ADS  CAS  Article  Google Scholar 

  38. Thevenaz, P., Ruttimann, U. E. & Unser, M. A pyramid approach to subpixel registration based on intensity. IEEE Trans. Image Process. 7, 27–41 (1998)

    ADS  CAS  Article  Google Scholar 

  39. Tian, L. et al. Imaging neural activity in worms, flies and mice with improved GCaMP calcium indicators. Nature Methods 6, 875–881 (2009)

    CAS  Article  Google Scholar 

Download references


We thank A. Kim and S. Kalina for technical assistance; L. L. Looger, J. Akerboom, D. S. Kim and the GENIE Project at Janelia Farm for making GCaMP available; E. Kyubwa and J. Keller for help with task development; A. D. Lien and M. Caudill for help with two-photon guided recordings; and M. Long, R. Malinow, G. Murphy, M. Scanziani and members of the Komiyama laboratory for comments and discussions. This work was supported by grants from Japan Science and Technology Agency (PRESTO), Pew Charitable Trusts, Alfred P. Sloan Foundation, David & Lucile Packard Foundation, Human Frontier Science Program and New York Stem Cell Foundation to T.K. A.J.P. is supported by the Neuroplasticity of Aging Training Grant (AG000216). S.X.C. is a Human Frontier Science Program postdoctoral fellow. T.K. is a NYSCF-Robertson Investigator.

Author information

Authors and Affiliations



A.J.P. and T.K. conceived the project. Dendritic spine imaging and optogenetic silencing experiments were performed by S.X.C. and analysed by S.X.C. and T.K. All other experiments were performed by A.J.P. and analysed by A.J.P. and T.K. A.J.P. and T.K. wrote the manuscript with input from S.X.C.

Corresponding author

Correspondence to Takaki Komiyama.

Ethics declarations

Competing interests

The authors declare no competing financial interests.

Extended data figures and tables

Extended Data Figure 1 Behaviour.

The fraction of rewarded trials is consistently high but the timing of behaviour improves during learning. a, Fraction of trials that are rewarded. b, Time from cue onset to movement onset decreases (P < 0.001, one-way ANOVA); inset, zoom. c, Time from movement onset to reward decreases (P < 0.001, one-way ANOVA); inset, zoom. d, The duration of each rewarded movement is stable throughout learning (P = 0.94, one-way ANOVA). Grey, individual mice; red, mean of all animals (a) or median of all trials (bd).

Extended Data Figure 2 Motor cortex is required for the lever-press task.

a, Aspiration lesion of motor cortex impairs learning. Mice were allowed to recover for 14 days after lesion before training. Left: histological image showing lesion in the right motor cortex and quantification of lesion extents in four mice shown as a density map of the fraction of animals in which the area was lesioned. Anterior is to the top; lateral to the right. + denotes bregma. The white circle indicates the imaged area. Middle: average time from movement onset to reward throughout learning. This time is longer in mice with motor cortex lesion (P < 0.01, two-way ANOVA), indicating that the mice with a lesion are less efficient in their movements. Right: correlation of lever movements in all pairs of trials within each session throughout learning. This correlation is lower in the mice with a lesion (P < 0.001, two-way ANOVA), indicating that the mice with a lesion do not develop reproducible movements. b, Injections of muscimol, a GABA receptor agonist, into the imaged area acutely impairs performance (control versus muscimol in motor cortex, **P < 0.01, Wilcoxon rank sum test). Muscimol injections in the barrel cortex had no significant effect (control versus muscimol in barrel cortex, P = 0.35, Wilcoxon rank sum test). Control, n = 18 sessions in 6 mice; barrel cortex, n = 6 sessions in 6 mice; motor cortex, n = 6 sessions in 6 mice. c, The imaged cortical area was acutely inactivated by stimulation of ChR2 in parvalbumin-expressing inhibitory neurons by blue light in interleaved 20% of trials (n = 10 sessions in 2 animals). This optogenetic inactivation of the imaged area impaired performance on a trial-by-trial basis (***P < 0.001, Wilcoxon rank sum test). Blue light had no effect on behaviour when the window was covered with opaque silicone (n = 4 sessions in 2 animals). All error bars are s.e.m.

Extended Data Figure 3 Optogenetic stimulation of the imaged area evokes forelimb movements in awake mice.

a, Optogenetic excitation of the imaged area triggers forelimb movements in mice expressing ChR2 but not in control mice not expressing ChR2 (P < 0.001, chi-squared test). ChR2 expression does not alter spontaneous movement frequency in the absence of stimulation (P = 0.64, chi-squared test). ‘During light’, 1-s light stimulation; ‘before light’, 2 to 1 s before light onset, n = 40 ‘during light’ trials and 38 ‘before light’ trials in two ChR2 mice, 38 ‘during light’ trials and 38 ‘before light’ trials in two control mice. b, Histological section showing the expression of ChR2 in the motor cortex. Green, ChR2–YFP; blue, DAPI.

Extended Data Figure 4 Simultaneous cell-attached recordings and two-photon calcium imaging in awake mice.

a, Left: in vivo two-photon image of motor cortex neurons expressing GCaMP5G. The neuron in the centre is targeted with a patch electrode. Right: after the recording session, voltage step was applied to the electrode to activate the recorded neuron. The increased GCaMP5G fluorescence in the middle neuron confirms that the neuron was indeed targeted. b, Example GCaMP5G fluorescence trace (top: black indicates fluorescence trace and red indicates detected calcium events) and simultaneously recorded action potentials (bottom: black vertical ticks; the numbers indicate the number of action potentials contained in each burst). Horizontal red lines at bottom indicate the duration of detected calcium events. Note the precise temporal relationship between action potentials and calcium events. c, Table summarizing data from six neurons in two mice. Positive offsets indicate the lag of the onset of detected calcium events relative to the first spike in the burst. The offset (7.1 ± 41.4 ms) is on the order of the temporal resolution of our imaging (35 ms per image frame).

Extended Data Figure 5 Lack of spatial clustering of movement-related excitatory neurons.

Each plot represents one animal. Red dots show the mean pairwise distance between movement-related excitatory neurons. Solid and dotted black lines show the mean and 95% confidence intervals, respectively, obtained from shuffling the identities of movement-related neurons among all excitatory neurons 10,000 times. Dots below the lower dotted line would indicate significant clustering of cells, whereas dots above the upper dotted line would indicate the significant dispersion of cells (P < 0.05).

Extended Data Figure 6 Additional analysis of population activity.

a, Cumulative distribution of fraction of sessions classified as movement-related for inhibitory (red) and excitatory (green) neurons, showing the relative invariance of inhibitory neurons and dynamism of excitatory neurons (P < 0.001, Kolmogorov–Smirnov test). b, Movement-related excitatory neuron populations in each session compared to the previous session. Grey, fraction of neurons classified in the previous session; white, not classified in the previous session. A large number of newly movement-related neurons were added in the first few sessions (P < 0.001, comparison between sessions 2–4 versus 10–14, Wilcoxon rank sum test). c, Fraction of excitatory neurons classified as movement-related in each session. Black, training (n = 7 mice, this is the data shown in Fig. 2b); red, no training (n = 6 mice). The expansion of movement-related neurons is specific to animals that underwent training (P = 0.74, sessions 1–2 combined; P < 0.001, sessions 3–7 combined; Wilcoxon rank sum test). d, Average population activity aligned to movement onset (black dotted line). Average activity (calcium event trace) of each movement-related excitatory neuron was averaged. The population activity diverged from baseline 105 ms before movement onset (red dotted line, Methods). e, Standard deviation of activity timing of individual movement-related excitatory neurons across sessions. Focusing on neurons that are classified as movement-related in three or more sessions, the standard deviation of activity onset timing relative to movement onset is plotted across sessions. Sessions were binned into one-third of the total number of sessions each neuron was classified. Activity timing became more stable on the neuron-by-neuron basis (r = −0.14, P < 0.001). f, Histogram of the time from movement onset that the activity of each movement-related neuron significantly diverged from baseline. 9.2% of movement-related excitatory neurons show significant pre-movement activity, a composition similar to a previous study14. 82.7% of activity of movement-related neurons occurred during the periods between 105 ms before movement onset and movement offset (Methods). g, The cumulative fraction plot of the timing of all activity onsets of movement-related excitatory neurons during rewarded movements. Each group of sessions is shown as a line, with different colours representing different sessions. The distribution of activity onset timing during later sessions shifts towards the movement onset (P < 0.001, Kolmogorov–Smirnov test for all three comparisons). All error bars are s.e.m.

Extended Data Figure 7 Activity analysis focusing on the first 500 ms of each movement.

For the activity analyses in the main figures that used the duration of 3 s after movement onset, we repeated the same analyses focusing on the first 500 ms of each movement (median time from movement onset to reward = 506 ms). This early activity shows progression throughout learning, similar to when activity over 3 s was considered. a, Standard deviation of the timing of activity onsets for movement-related excitatory neurons over sessions, indicating a gradual refinement of activity timing (r = −0.18, P < 0.001). Neurons that were active in less than five trials of a given session were excluded from this analysis. The first bin contains only one data point and thus does not have an error bar. This analysis is equivalent to Fig. 2g. b, Pairwise trial-to-trial correlation of temporal population activity vectors increases with learning (r = 0.38, P < 0.001). Temporal population activity vector was defined as a concatenation of the activity traces of all movement-related neurons and thus maintained temporal information within each movement. This analysis is equivalent to Fig. 2h. c, Correlation of spatiotemporal activity with the learned activity pattern is a function of the correlation of movement with the learned movement pattern in expert sessions. Movements similar to the learned movement pattern but made in naive sessions display activity very different from the learned activity pattern (P = 0.28 and <0.001 in the bins 1 and 2–10, respectively, Wilcoxon rank sum test). This analysis is equivalent to Fig. 3b. d, Pairwise trial-to-trial correlation of temporal population activity vectors plotted as a function of movement correlation on those trials. A strong relationship between population activity and movement emerges during learning (P = 0.08, = 0.08, = 0.004, <0.001, = 0.002, <0.001, = 0.001, = 0.002, <0.001 and = 0.046 for each bin, Wilcoxon rank sum test). This analysis is equivalent to Fig. 3c. All error bars are s.e.m.

Extended Data Figure 8 Dynamics of dendritic spines in the hindlimb area during learning of the lever-press task.

Summary of dendritic spine dynamics in the hindlimb area during control period (7 days before training) and subsequent 7 days of training. Mice were water restricted in both conditions. Top: spine additions (black) and eliminations (grey) in each session. For control sessions, data from all sessions are combined. Bottom: total spine number across sessions. Values are normalized to the total spine number in session 1 in each condition. Unlike the forelimb area, the density of dendritic spines in the hindlimb area is relatively stable during learning (P = 0.07, comparisons between control versus training sessions 4–7, Wilcoxon rank sum test). All error bars are s.e.m.

Extended Data Figure 9 Schematic of learning-related changes in the relationship of motor cortex activity and movement.

Top: abstract space of activity patterns. Bottom: abstract space of movements. Circles in the movement space represent observed movements, and ovals in the activity space represent possible activity patterns that can lead to corresponding movements. Crosses and arrows represent example individual trials of activity–movement pairs. In naive animals, each trial involves variable activity and movement patterns as illustrated by scattered crosses and multiple movements. In this stage, the relationship between activity and movement is inconsistent (that is, degenerate), such that same movement is derived from different activity patterns in different trials. During learning, this degeneracy is reduced and a reproducible spatiotemporal activity pattern emerges in the motor cortex that reliably generates the learned movement. This learned activity pattern (bold cross) is rarely, if at all, observed in naive stages.

PowerPoint slides

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Peters, A., Chen, S. & Komiyama, T. Emergence of reproducible spatiotemporal activity during motor learning. Nature 510, 263–267 (2014).

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI:

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.


Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing