Skip to main content

Thank you for visiting nature.com. 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.

  • Article
  • Published:

High-order coordination of cortical spiking activity modulates perceptual accuracy

Nature Neuroscience volume 22, pages 1148–1158 (2019)Cite this article

Subjects

Abstract

The accurate relay of electrical signals within cortical networks is key to perception and cognitive function. Theoretically, it has long been proposed that temporal coordination of neuronal spiking activity controls signal transmission and behavior. However, whether and how temporally precise neuronal coordination in population activity influences perception are unknown. Here, we recorded populations of neurons in early and mid-level visual cortex (areas V1 and V4) simultaneously to discover that the precise temporal coordination between the spiking activity of three or more cells carries information about visual perception in the absence of firing rate modulation. The accuracy of perceptual responses correlated with high-order spiking coordination within V4, but not V1, and with feedforward coordination between V1 and V4. These results indicate that while visual stimuli are encoded in the discharge rates of neurons, perceptual accuracy is related to temporally precise spiking coordination within and between cortical networks.

This is a preview of subscription content, access via your institution

Access options

Buy this article

  • Purchase on Springer Link
  • Instant access to full article PDF
Buy now

Prices may be subject to local taxes which are calculated during checkout

Fig. 1: High-order coordination in spiking activity during a behavioral task.
Fig. 2: Simulation of correlated spike trains with statistics similar to those from recorded neurons, except for coordinated spiking.
Fig. 3: Relationship between high-order coordination rate and perceptual accuracy.
Fig. 4: Population effects: correct trials are associated with higher order coordination in area V4.
Fig. 5: Feedforward coordination between V1 and V4 neurons is related to perceptual accuracy.
Fig. 6: Coordination rates are unrelated to eye movements.

Similar content being viewed by others

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Code availability

The custom-written software supporting the findings of this study is available from the corresponding author upon reasonable request.

References

  1. Averbeck, B. B., Latham, P. E. & Pouget, A. Neural correlations, population coding and computation. Nat. Rev. Neurosci. 7, 358–366 (2006).

    Article  CAS  Google Scholar 

  2. Shadlen, M. N. & Newsome, W. T. The variable discharge of cortical neurons: implications for connectivity, computation, and information coding. J. Neurosci. 18, 3870–3896 (1998).

    Article  CAS  Google Scholar 

  3. Zandvakili, A. & Kohn, A. Coordinated neuronal activity enhances corticocortical communication. Neuron 87, 827–839 (2015).

    Article  CAS  Google Scholar 

  4. Heinzle, J., König, P. & Salazar, R. F. Modulation of synchrony without changes in firing rates. Cogn. Neurodyn. 1, 225–235 (2007).

    Article  Google Scholar 

  5. Trousdale, J., Hu, Y., Shea-Brown, E. & Josić, K. A generative spike train model with time-structured higher order correlations. Front. Comput. Neurosci. 7, 84 (2013).

    Article  Google Scholar 

  6. Abeles, M. Corticonics: Neural Circuits of the Cerebral Cortex. (Cambridge Univ. Press, 1991).

  7. Dan, Y., Alonso, J. M., Usrey, W. M. & Reid, R. C. Coding of visual information by precisely correlated spikes in the lateral geniculate nucleus. Nat. Neurosci. 1, 501–507 (1998).

    Article  CAS  Google Scholar 

  8. Takeuchi, D., Hirabayashi, T., Tamura, K. & Miyashita, Y. Reversal of interlaminar signal between sensory and memory processing in monkey temporal cortex. Science 331, 1443–1447 (2011).

    Article  CAS  Google Scholar 

  9. Uhlhaas, P. et al. Neural synchrony in cortical networks: history, concept and current status. Front. Integr. Neurosci. 3, 17 (2009).

    Article  Google Scholar 

  10. Singer, W. & Gray, C. M. Visual feature integration and the temporal correlation hypothesis. Annu. Rev. Neurosci. 18, 555–586 (1995).

    Article  CAS  Google Scholar 

  11. Histed, M. H. & Maunsell, J. H. R. Cortical neural populations can guide behavior by integrating inputs linearly, independent of synchrony. Proc. Natl Acad. Sci. USA 111, E178–E187 (2013).

    Article  Google Scholar 

  12. Alonso, J. M., Usrey, W. M. & Reid, R. C. Precisely correlated firing in cells of the lateral geniculate nucleus. Nature 383, 815–819 (1996).

    Article  CAS  Google Scholar 

  13. Bruno, R. M. & Sakmann, B. Cortex is driven by weak but synchronously active thalamocortical synapses. Science 312, 1622–1627 (2006).

    Article  CAS  Google Scholar 

  14. Salinas, E. & Sejnowski, T. J. Impact of correlated synaptic input on output firing rate and variability in simple neuronal models. J. Neurosci. 20, 6193–6209 (2000).

    Article  CAS  Google Scholar 

  15. Renart, A. et al. The asynchronous state in cortical circuits. Science 327, 587–590 (2010).

    Article  CAS  Google Scholar 

  16. Wehr, M. & Zador, A. M. Balanced inhibition underlies tuning and sharpens spike timing in auditory cortex. Nature 426, 442–446 (2003).

    Article  CAS  Google Scholar 

  17. Gabernet, L., Jadhav, S. P., Feldman, D. E., Carandini, M. & Scanziani, M. Somatosensory integration controlled by dynamic thalamocortical feed-forward inhibition. Neuron 48, 315–327 (2005).

    Article  CAS  Google Scholar 

  18. Fries, P. Neuronal gamma-band synchronization as a fundamental process in cortical computation. Annu. Rev. Neurosci. 32, 209–224 (2009).

    Article  CAS  Google Scholar 

  19. Ecker, A. S. et al. State dependence of noise correlations in macaque primary visual cortex. Neuron 82, 235–248 (2014).

    Article  CAS  Google Scholar 

  20. Salazar, R. F., Dotson, N. M., Bressler, S. L. & Gray, C. M. Content-specific fronto-parietal synchronization during visual working memory. Science 338, 1097–1101 (2012).

    Article  CAS  Google Scholar 

  21. Bosman, C. A. et al. Attentional stimulus selection through selective synchronization between monkey visual areas. Neuron 75, 875–888 (2012).

    Article  CAS  Google Scholar 

  22. van Kerkoerle, T. et al. Alpha and gamma oscillations characterize feedback and feedforward processing in monkey visual cortex. Proc. Natl Acad. Sci. USA 111, 14332–14341 (2014).

    Article  Google Scholar 

  23. Jia, X., Smith, M. A. & Kohn, A. Stimulus selectivity and spatial coherence of gamma components of the local field potential. J. Neurosci. 31, 9390–9403 (2011).

    Article  CAS  Google Scholar 

  24. Schroeder, C. E., Mehta, A. D. & Givre, S. J. A spatiotemporal profile of visual system activation revealed by current source density analysis in the awake macaque. Cereb. Cortex 8, 575–592 (1998).

    Article  CAS  Google Scholar 

  25. Hansen, B. J., Chelaru, M. I. & Dragoi, V. Correlated variability in laminar cortical circuits. Neuron 76, 590–602 (2012).

    Article  CAS  Google Scholar 

  26. Pipa, G., Wheeler, D. W., Singer, W. & Nikolić, D. NeuroXidence: reliable and efficient analysis of an excess or deficiency of joint-spike events. J. Comput. Neurosci. 25, 64–88 (2008).

    Article  Google Scholar 

  27. Schneidman, E., Berry, M. J., Segev, R. & Bialek, W. Weak pairwise correlations imply strongly correlated network states in a neural population. Nature 440, 1007–1012 (2006).

    Article  CAS  Google Scholar 

  28. Benjamini, Y. & Hochberg, Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J. R. Stat. Soc. Ser. B 57, 289–300 (1995).

    Google Scholar 

  29. Britten, K. H., Newsome, W. T., Shadlen, M. N., Celebrini, S. & Movshon, J. A. A relationship between behavioral choice and the visual responses of neurons in macaque MT. Vis. Neurosci. 13, 87 (2009).

    Article  Google Scholar 

  30. Beaman, C. B., Eagleman, S. L. & Dragoi, V. Sensory coding accuracy and perceptual performance are improved during the desynchronized cortical state. Nat. Commun. 8, 1–14 (2017).

    Article  CAS  Google Scholar 

  31. Gutnisky, D. A., Beaman, C., Lew, S. E. & Dragoi, V. Cortical response states for enhanced sensory discrimination. eLife 6, e29226 (2017).

    Article  Google Scholar 

  32. Cohen, M. R. & Maunsell, J. H. R. Attention improves performance primarily by reducing interneuronal correlations. Nat. Neurosci. 12, 1594–1601 (2009).

    Article  CAS  Google Scholar 

  33. McAdams, C. J. & Maunsell, J. H. R. Effects of attention on the reliability of individual neurons in monkey visual cortex. Neuron 23, 765–773 (1999).

    Article  CAS  Google Scholar 

  34. Bair, W., Zohary, E. & Newsome, W. T. Correlated firing in macaque visual area MT: time scales and relationship to behavior. J. Neurosci. 21, 1676–1697 (2001).

    Article  CAS  Google Scholar 

  35. Emmerich, D. S. Signal Detection Theory and Psychophysics by David M. Green, John A. Swets. Q. Rev. Biol. 42, 578 (1967).

    Article  Google Scholar 

  36. Ungerleider, L. G., Galkin, T. W., Desimone, R. & Gattass, R. Cortical connections of area V4 in the macaque. Cereb. Cortex 18, 477–499 (2008).

    Article  Google Scholar 

  37. Markov, N. T. et al. Anatomy of hierarchy: feedforward and feedback pathways in macaque visual cortex. J. Comp. Neurol. 522, 225–259 (2014).

    Article  Google Scholar 

  38. Lee, J., Williford, T. & Maunsell, J. H. Spatial attention and the latency of neuronal responses in macaque area V4. J. Neurosci. 27, 9632–9637 (2007).

    Article  CAS  Google Scholar 

  39. Gollisch, T. & Meister, M. Rapid neural coding in the retina with relative spike latencies. Science 319, 1108–1111 (2008).

    Article  CAS  Google Scholar 

  40. Dan, Y., Atick, J. J. & Reid, R. C. Efficient coding of natural scenes in the lateral geniculate nucleus: experimental test of a computational theory. J. Neurosci. 16, 3351–3362 (1996).

    Article  CAS  Google Scholar 

  41. Hirabayashi, T. & Miyashita, Y. Dynamically modulated spike correlation in monkey inferior temporal cortex depending on the feature configuration within a whole object. J. Neurosci. 25, 10299–10307 (2005).

    Article  CAS  Google Scholar 

  42. Vaadia, E. et al. Dynamics of neuronal interactions in monkey cortex in relation to behavioural events. Nature 373, 515–518 (1995).

    Article  CAS  Google Scholar 

  43. Gilbert, C. D. & Li, W. Top-down influences on visual processing. Nat. Rev. Neurosci. 14, 350–363 (2013).

    Article  CAS  Google Scholar 

  44. Smolyanskaya, A., Haefner, R. M., Lomber, S. G. & Born, R. T. A modality-specific feedforward component of choice-related activity in MT. Neuron 87, 208–219 (2015).

    Article  CAS  Google Scholar 

  45. Crick, F. & Koch, C. Constraints on cortical and thalamic projections: the no-strong-loops hypothesis. Nature 391, 245–250 (1998).

    Article  CAS  Google Scholar 

  46. Felleman, D. J. & Van Essen, D. C. Distributed hierarchical processing in the primate cerebral cortex. Cereb. Cortex. 1, 1–47 (1991).

    Article  CAS  Google Scholar 

  47. Fries, P. A mechanism for cognitive dynamics: neuronal communication through neuronal coherence. Trends Cogn. Sci. 9, 474–480 (2005).

    Article  Google Scholar 

  48. Okun, M. et al. Population rate dynamics and multineuron firing patterns in sensory cortex. J. Neurosci. 32, 17108–17119 (2012).

    Article  CAS  Google Scholar 

  49. Riehle, A. Spike synchronization and rate modulation differentially involved in motor cortical function. Science 278, 1950–1953 (1997).

    Article  CAS  Google Scholar 

  50. Yang, Y. & Zador, A. M. Differences in sensitivity to neural timing among cortical areas. J. Neurosci. 32, 15142–15147 (2012).

    Article  CAS  Google Scholar 

  51. Bishop C. M. Pattern Recognition and Machine Learning (Springer, 2006).

  52. N. Haidous, N. & S. Sawilowsky, S. Robustness and power of the Kornbrot rank difference, signed ranks, and dependent samples T-test. Am. J. Appl. Math. Stat. 1, 99–102 (2013).

    Article  Google Scholar 

  53. Sawilowsky, S. S. & Blair, R. C. A more realistic look at the robustness and Type II error properties of the t test to departures from population normality. Psychol. Bull. 111, 352–360 (1992).

    Article  Google Scholar 

Download references

Acknowledgements

The authors thank X. Pitkow, H. Shouval, and J. Magnotti for discussions and comments. This work was supported by NIH EUREKA and NEI grants to V.D.

Author information

Author notes

  1. Neda Shahidi

    Present address: Department of Ophthalmology, Universitätsmedizin Göttingen, Göttingen, Germany

  2. Ming Hu

    Present address: Department of Brain and Cognitive Sciences, Massachusetts Institute of Technology, Cambridge, MA, USA

Authors and Affiliations

  1. Department of Neurobiology and Anatomy, McGovern Medical School, University of Texas, Houston, TX, USA

    Neda Shahidi, Ariana R. Andrei, Ming Hu & Valentin Dragoi

Authors
  1. Neda Shahidi
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Ariana R. Andrei
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Ming Hu
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Valentin Dragoi
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

A.R.A., M.H., and V.D. designed the experiments. A.R.A. and M.H. performed the experiments. N.S. analyzed the data with guidance from V.D. N.S. and V.D. wrote the manuscript.

Corresponding author

Correspondence to Valentin Dragoi.

Ethics declarations

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

These authors contributed equally: Ariana R. Andrei, Ming Hu.

Integrated supplementary information

Supplementary Fig. 1 Example cross-correlograms between neurons in a triplet.

We calculated the coordination rates using conventional cross-correlograms8 between all possible pairs of three participating neurons (neurons 5, 6, and 7) for an example triplet. No significant difference between the peaks of cross-correlograms was observed (p > 0.1, 2-sided unpaired t-test). Shaded regions represent SEM across trials (correct: n = 69, incorrect: n = 31). Across all sessions (V1: n = 22, V4: n = 12), the difference between correct and incorrect CCG peaks was not statistically significant (2-sided Wilcoxon signed rank test with FDR correction; V1: p = 0.27, V4: p = 0.79).

Supplementary Fig. 2 Coordination rates for V1 and V4.

Normalized coordination rates for cell ensembles of size 2 to 5 in areas V1 (22 sessions) and V4 (12 sessions). Error bars represent SEM.

Supplementary Fig. 3 Stimulus classification performance in areas V1 and V4 of a SVM decoder with RBF kernel.

The inputs to the decoder were the coordination rates of pairs, triplets, and quarters in each area. The decoder’s output is the classification of test stimulus identity (stim1 vs. stim2). The analysis window is a 300-ms window of the sustained response to the test stimulus (150–450 ms relative to stimulus onset). The analysis window is the 300-ms period of the sustained response to the test stimulus (150–450 ms relative to stimulus onset; this window provides the most significant results in the analysis of coordination rates in relation to perceptual accuracy, MEP_L_fig3Fig. 3; choosing a window including the onset transient response yields similar results). The parameters of the cross-validation are stated in the original manuscript. The data points represent averages and error bars represent SEM across sessions (n = 22 for V1 and n = 12 for V4; decoder performance was not statistically significant in any session regardless of the cortical area that was tested; p > 0.1, 2-sided Wilcoxon rank-sum test with FDR correction).

Supplementary Fig. 4 The main results of our study shown for individual animals.

a) Performance of stimulus decoder tested for correct and incorrect trials for a 200-ms analysis window (to maximize decoder performance), as in Fig. 3a (asterisks indicate p < 0.01). This figure confirms that (in each animal) stimuli can be decoded above chance level both in correct and incorrect trials in V1, but only in correct trials in V4. The bars show the average across sessions (monkey W: n = 8 in V1, and n = 8 in V4; monkey C: n = 14 in V1, and n = 4 in V4). Error bars represent SEM. b) Normalized coordination rates (correct - incorrect trials) in V4 for 2nd and higher order events during the 300-ms window associated with the target (left) and test (right) stimuli, as well as the delay period (100–400 ms following target onset). As in the original average results across animals, the only significant cases are the occurrence of higher order coordinated events during the test stimulus presentation (asterisks indicate p < 0.01; the actual p-values are reported in the manuscript). The bars show the average across sessions (monkey W: n = 8 for SO and n = 16 for HO; monkey C: n = 4 for SO and n = 8 for HO). Error bars show SEM. c) Peak coordination rate between V1 and V4 for correct and incorrect trials. For all higher order events combined, the p-value is 0.0029 for monkey W (n = 16) and 0.0004 for monkey C (n = 16). The traces represent the average and the error-bars represent SEM. For the high-order cross-correlation analysis (manuscript MEP_L_fig5Fig. 5e), the p-values are 0.0420 for monkey W (n = 4) and 0.0117 for monkey C (n = 4). This figure confirms that (for each animal) there is higher coordination of high-order spiking events in correct trials compared to incorrect.

Supplementary Fig. 5 Firing rates of neurons do not differ between correct and incorrect trials.

a) Distribution of firing rates of a sample neuron calculated for 20 ms time bins, separately for a correct and incorrect trial. b) Percentage of neurons (n = 293) for which the distributions of firing rates for correct and incorrect trials were statistically different (2-sided Wilcoxon rank-sum test, p < 0.05 with FDR correction) as a function of bin size (by pooling cells across sessions). The two firing rate distributions were constructed by pooling the firing rates of a neuron across trials for each time bin, separately for correct and incorrect trials. Subsequently, for each time bin we tested whether the two response distributions are significantly different. This analysis was repeated for each cell in the population. c,d) Baseline-removed population-averaged PSTHs of all the neurons in V1 (n = 183) and V4 (n = 110) for correct and incorrect trials separately for the match conditions (c) and non-match condition in each monkey (d). The bin size was 20 ms. (2-sided Wilcoxon rank-sum test with FDR correction; match condition: p = 0.08; monkey W, p = 0.51 for V1 and p = 0.16 for V4. monkey C: p = 0.22 for V1 and p = 0.19 for V4). Shaded regions show SEM.

Supplementary Fig. 6 Performance of the classifier that uses the firing rates of all simultaneously recorded neurons to decode perceptual accuracy (correct vs incorrect).

A 200-ms window sliding in 20-ms incremental steps was used. The classifier was trained using 80% of the trials and performance was measured by classifying the remaining 20% of trials. The performance was cross-validated 1,000 times by randomly dividing trials to training and test sets. In shuffled classifiers, the trial labels (correct vs. incorrect) were shuffled before training; then the cross-validation procedure was repeated. The performance of the classifier was not significantly different from the null hypothesis (p > 0.1) for any analysis window in any cortical area. Shaded regions represent SEM across sessions (n = 22 for V1 and n = 12 for V4).

Supplementary Fig. 7 Three measures of neural activity in V1 and V4 related to attentional modulation.

Mean firing rates (panel a), Fano factor (panel b), and mean noise correlations (panel c), previously demonstrated to be correlated with attentional modulation, do not exhibit statistically significant differences between correct and incorrect trials. The vertical lines in each panel represent the means of each distribution. All tests were 2-sided; V1: n = 183, V4: n = 110.

Supplementary Fig. 8 Differential coordination rates for the sample session in Fig. 3.

The difference between coordination rates of correct and incorrect trials are shown separately for second-order and higher-order coordinated. Coordination rates were calculated for a 300 ms window sliding in 10 ms steps over the test stimulus. Each trace represents the averages across trials (correct: n = 46, incorrect: n = 46); shaded areas represent SEM.

Supplementary Fig. 9 Differential coordination rates for the ‘Target’ and ‘Delay’ components of a trial.

Distribution of the difference in coordination rates between correct and incorrect trials for 12 V4 sessions during the target (left) and delay (right) intervals. The windows of analysis are identical to those in MEP_L_fig4Fig. 4b. All analysis parameters are the same as in Fig. 4b.

Supplementary Fig. 10 Trial-shuffled coordination rates.

As an additional control, we shuffled the spike trains associated with correct and incorrect trial sets (within each set), then recalculated the coordination rates for pairs, triplets, and quartets, and finally subtracted the coordination rates for the incorrect trials from the rates associated with correct trials. No significant difference between coordination rates was observed (p > 0.1) when spike trains were shuffled. Horizontal lines show the average across sessions (V1: n = 22, V4: n = 12).

Supplementary Fig. 11 Coordination rates are significantly different between correct and incorrect trials (p < 0.05) for a range of bin sizes and jitter ranges.

a) Coordination rates as a function of bin size (top) and associated p-values (bottom), 2-sided Wilcoxon signed rank test, calculated across V4 sessions (n = 12). The bin width varied in the range of 1 and 11 ms with steps of 1 ms. For bin sizes of 4 ms and higher, p-values of triplets and quartets are smaller than 0.05. b) Same analysis as in (a) but for jitter range of 3 to 13 ms with steps of 1 ms. For jitter range of 8 ms and higher, the p-value was smaller than 0.05.

Supplementary Fig. 12 Cross-correlogram (CCG) between V1 and V4 averaged across pairs of cells.

Each cross-correlogram was normalized by the geometric mean of the firing rates of participating neurons. No significant peak in the CCG (z-score > 2) was observed within the ± 40 ms time lag range. Shaded area represents SEM across all V1-V4 pairs (n = 734).

Supplementary Fig. 13 Cross-correlogram (CCG) between V1 and V4 to test whether high-order spiking coordination drives spiking activity in a target area.

Left: CCG between convergent feedforward inputs from V1 (high-order spiking events) and V4 individual spikes reveals that high-order coordination in V1 is not associated with elevated spiking in V4. Right: CCG between convergent feedback inputs from V4 (high-order spiking events) and V1 individual spikes reveals that high-order coordination in V4 is not associated with elevated spiking in V1. Each cross-correlogram was normalized by the geometric mean of the firing rates of participating neurons. No significant CCG peak (z-score > 2) was observed within the ± 40 ms time lag range that we explored. Shaded areas represent SEM across sessions (n = 7).

Supplementary Fig. 14 Cross-correlogram between higher-order events in V1 and individual neurons in V4, sorted by the magnitude of their peak.

Each row shows the trial averaged cross-correlogram for one V4 neuron. The peaks were z-scored using the tail of the cross-correlogram, similar to the original analysis in the manuscript; only 3 cross-correlograms showed a significant peak (p < 0.05, t-test).

Supplementary Fig. 15 Correlation between ‘local’ behavioral performance and coordination rates.

a) ‘Local’ behavioral performance was calculated as percentage of correct responses in the last 10 trials for a sample session. In the same panel we represent the coordination rate in the trial immediately following the window of trials for which behavioral performance was calculated. The Pearson’s correlation coefficient between local behavioral performance and coordination rate and the associated p-value are shown (2-sided t-test was used to calculate p-value). b) Pearson’s correlation coefficient of coordination rates with behavioral performance in the recent trials. The width of the window for which behavioral performance was measured is 1–30 trials before current trial. The Pearson’s correlation coefficient between the coordination rates and behavioral outcomes in the current trial (0 for incorrect and 1 for correct) is shown as a separate data point. The red traces and dots represent averages across V4 sessions (n = 12); error-bars and shaded regions represent SEM.

Supplementary information

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Shahidi, N., Andrei, A.R., Hu, M. et al. High-order coordination of cortical spiking activity modulates perceptual accuracy. Nat Neurosci 22, 1148–1158 (2019). https://doi.org/10.1038/s41593-019-0406-3

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1038/s41593-019-0406-3

This article is cited by

Search

Advanced search

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