Expectation-induced modulation of metastable activity underlies faster coding of sensory stimuli


Sensory stimuli can be recognized more rapidly when they are expected. This phenomenon depends on expectation affecting the cortical processing of sensory information. However, the mechanisms responsible for the effects of expectation on sensory circuits remain elusive. In the present study, we report a novel computational mechanism underlying the expectation-dependent acceleration of coding observed in the gustatory cortex of alert rats. We use a recurrent spiking network model with a clustered architecture capturing essential features of cortical activity, such as its intrinsically generated metastable dynamics. Relying on network theory and computer simulations, we propose that expectation exerts its function by modulating the intrinsically generated dynamics preceding taste delivery. Our model’s predictions were confirmed in the experimental data, demonstrating how the modulation of ongoing activity can shape sensory coding. Altogether, these results provide a biologically plausible theory of expectation and ascribe an alternative functional role to intrinsically generated, metastable activity.

Access options

Rent or Buy article

Get time limited or full article access on ReadCube.


All prices are NET prices.

Fig. 1: Anticipatory activity requires a clustered network architecture.
Fig. 2: Robustness of anticipatory activity to variations in stimulus and cue models.
Fig. 3: Anticipatory cue speeds up network dynamics.
Fig. 4: Anticipatory cue induces faster onset of stimulus-coding states.
Fig. 5: Anticipation of coding states: model versus data.

Data availability

Experimental datasets are available from the authors on request.

Code availability

All data analysis and network simulation scripts are available from the authors on request. A demo code for simulating the network model is available on GitHub (https://github.com/mazzulab).


  1. 1.

    Gilbert, C. D. & Sigman, M. Brain states: top-down influences in sensory processing. Neuron 54, 677–696 (2007).

    CAS  Article  Google Scholar 

  2. 2.

    Jaramillo, S. & Zador, A. M. The auditory cortex mediates the perceptual effects of acoustic temporal expectation. Nat. Neurosci. 14, 246–251 (2011).

    CAS  Article  Google Scholar 

  3. 3.

    Engel, A. K., Fries, P. & Singer, W. Dynamic predictions: oscillations and synchrony in top-down processing. Nat. Rev. Neurosci. 2, 704–716 (2001).

    CAS  Article  Google Scholar 

  4. 4.

    Doherty, J. R., Rao, A., Mesulam, M. M. & Nobre, A. C. Synergistic effect of combined temporal and spatial expectations on visual attention. J. Neurosci. 25, 8259–8266 (2005).

    CAS  Article  Google Scholar 

  5. 5.

    Niwa, M., Johnson, J. S., O’Connor, K. N. & Sutter, M. L. Active engagement improves primary auditory cortical neurons’ ability to discriminate temporal modulation. J. Neurosci. 32, 9323–9334 (2012).

    CAS  Article  Google Scholar 

  6. 6.

    Samuelsen, C. L., Gardner, M. P. & Fontanini, A. Effects of cue-triggered expectation on cortical processing of taste. Neuron 74, 410–422 (2012).

    CAS  Article  Google Scholar 

  7. 7.

    Yoshida, T. & Katz, D. B. Control of prestimulus activity related to improved sensory coding within a discrimination task. J. Neurosci. 31, 4101–4112 (2011).

    CAS  Article  Google Scholar 

  8. 8.

    Gardner, M. P. & Fontanini, A. Encoding and tracking of outcome-specific expectancy in the gustatory cortex of alert rats. J. Neurosci. 34, 13000–13017 (2014).

    CAS  Article  Google Scholar 

  9. 9.

    Vincis, R. & Fontanini, A. Associative learning changes cross-modal representations in the gustatory cortex. eLife 5, e16420 (2016).

    Article  Google Scholar 

  10. 10.

    Amit, D. J. & Brunel, N. Model of global spontaneous activity and local structured activity during delay periods in the cerebral cortex. Cereb. Cortex 7, 237–252 (1997).

    CAS  Article  Google Scholar 

  11. 11.

    Deco, G. & Hugues, E. Neural network mechanisms underlying stimulus driven variability reduction. PLoS Comput. Biol. 8, e1002395 (2012).

    CAS  Article  Google Scholar 

  12. 12.

    Harvey, C. D., Coen, P. & Tank, D. W. Choice-specific sequences in parietal cortex during a virtual-navigation decision task. Nature 484, 62–68 (2012).

    CAS  Article  Google Scholar 

  13. 13.

    Litwin-Kumar, A. & Doiron, B. Slow dynamics and high variability in balanced cortical networks with clustered connections. Nat. Neurosci. 15, 1498–1505 (2012).

    CAS  Article  Google Scholar 

  14. 14.

    Mazzucato, L., Fontanini, A. & La Camera, G. Dynamics of multistable states during ongoing and evoked cortical activity. J. Neurosci. 35, 8214–8231 (2015).

    CAS  Article  Google Scholar 

  15. 15.

    Mazzucato, L., Fontanini, A. & La Camera, G. Stimuli reduce the dimensionality of cortical activity. Front. Syst. Neurosci. 10, 11 (2016).

    Article  Google Scholar 

  16. 16.

    Abeles, M. et al. Cortical activity flips among quasi-stationary states. Proc. Natl Acad. Sci. USA 92, 8616–8620 (1995).

    CAS  Article  Google Scholar 

  17. 17.

    Seidemann, E., Meilijson, I., Abeles, M., Bergman, H. & Vaadia, E. Simultaneously recorded single units in the frontal cortex go through sequences of discrete and stable states in monkeys performing a delayed localization task. J. Neurosci. 16, 752–768 (1996).

    CAS  Article  Google Scholar 

  18. 18.

    Arieli, A., Shoham, D., Hildesheim, R. & Grinvald, A. Coherent spatiotemporal patterns of ongoing activity revealed by real-time optical imaging coupled with single-unit recording in the cat visual cortex. J. Neurophysiol. 73, 2072–2093 (1995).

    CAS  Article  Google Scholar 

  19. 19.

    Arieli, A., Sterkin, A., Grinvald, A. & Aertsen, A. Dynamics of ongoing activity: explanation of the large variability in evoked cortical responses. Science 273, 1868–1871 (1996).

    CAS  Article  Google Scholar 

  20. 20.

    Rich, E. L. & Wallis, J. D. Decoding subjective decisions from orbitofrontal cortex. Nat. Neurosci. 19, 973–980 (2016).

    CAS  Article  Google Scholar 

  21. 21.

    Ponce-Alvarez, A., Nácher, V., Luna, R., Riehle, A. & Romo, R. Dynamics of cortical neuronal ensembles transit from decision making to storage for later report. J. Neurosci. 32, 11956–11969 (2012).

    CAS  Article  Google Scholar 

  22. 22.

    Durstewitz, D., Vittoz, N. M., Floresco, S. B. & Seamans, J. K. Abrupt transitions between prefrontal neural ensemble states accompany behavioral transitions during rule learning. Neuron 66, 438–448 (2010).

    CAS  Article  Google Scholar 

  23. 23.

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

    CAS  Article  Google Scholar 

  24. 24.

    Chen, X., Gabitto, M., Peng, Y., Ryba, N. J. & Zuker, C. S. A gustotopic map of taste qualities in the mammalian brain. Science 333, 1262–1266 (2011).

    CAS  Article  Google Scholar 

  25. 25.

    Fletcher, M. L., Ogg, M. C., Lu, L., Ogg, R. J. & Boughter, J. D. Jr. Overlapping Representation of primary tastes in a defined region of the gustatory cortex. J. Neurosci. 37, 7595–7605 (2017).

    CAS  Article  Google Scholar 

  26. 26.

    Kiani, R. et al. Natural grouping of neural responses reveals spatially segregated clusters in prearcuate cortex. Neuron 85, 1359–1373 (2015).

    CAS  Article  Google Scholar 

  27. 27.

    Katz, D. B., Simon, S. A. & Nicolelis, M. A. Dynamic and multimodal responses of gustatory cortical neurons in awake rats. J. Neurosci. 21, 4478–4489 (2001).

    CAS  Article  Google Scholar 

  28. 28.

    Jezzini, A., Mazzucato, L., La Camera, G. & Fontanini, A. Processing of hedonic and chemosensory features of taste in medial prefrontal and insular networks. J. Neurosci. 33, 18966–18978 (2013).

    CAS  Article  Google Scholar 

  29. 29.

    van Vreeswijk, C. & Sompolinsky, H. Chaos in neuronal networks with balanced excitatory and inhibitory activity. Science 274, 1724–1726 (1996).

    Article  Google Scholar 

  30. 30.

    Curti, E., Mongillo, G., La Camera, G. & Amit, D. J. Mean field and capacity in realistic networks of spiking neurons storing sparsely coded random memories. Neural Comput. 16, 2597–2637 (2004).

    Article  Google Scholar 

  31. 31.

    Mascaro, M. & Amit, D. J. Effective neural response function for collective population states. Network 10, 351–373 (1999).

    CAS  Article  Google Scholar 

  32. 32.

    Mattia, M. et al. Heterogeneous attractor cell assemblies for motor planning in premotor cortex. J. Neurosci. 33, 11155–11168 (2013).

    CAS  Article  Google Scholar 

  33. 33.

    La Camera, G., Giugliano, M., Senn, W. & Fusi, S. The response of cortical neurons to in vivo-like input current: theory and experiment: I. Noisy inputs with stationary statistics. Biol. Cybern. 99, 279–301 (2008).

    Article  Google Scholar 

  34. 34.

    Hänggi, P., Talkner, P. & Borkovec, M. Reaction-rate theory: Fifty years after Kramers. Rev. Mod. Phys. 62, 251 (1990).

    Article  Google Scholar 

  35. 35.

    Kenet, T., Bibitchkov, D., Tsodyks, M., Grinvald, A. & Arieli, A. Spontaneously emerging cortical representations of visual attributes. Nature 425, 954–956 (2003).

    CAS  Article  Google Scholar 

  36. 36.

    Pastalkova, E., Itskov, V., Amarasingham, A. & Buzsáki, G. Internally generated cell assembly sequences in the rat hippocampus. Science 321, 1322–1327 (2008).

    CAS  Article  Google Scholar 

  37. 37.

    Luczak, A., Barthó, P. & Harris, K. D. Spontaneous events outline the realm of possible sensory responses in neocortical populations. Neuron 62, 413–425 (2009).

    CAS  Article  Google Scholar 

  38. 38.

    Puccini, G. D., Sanchez-Vives, M. V. & Compte, A. Integrated mechanisms of anticipation and rate-of-change computations in cortical circuits. PLoS Comput. Biol. 3, e82 (2007).

    Article  Google Scholar 

  39. 39.

    Miller, P. & Katz, D. B. Stochastic transitions between neural states in taste processing and decision-making. J. Neurosci. 30, 2559–2570 (2010).

    CAS  Article  Google Scholar 

  40. 40.

    Jones, L. M., Fontanini, A., Sadacca, B. F., Miller, P. & Katz, D. B. Natural stimuli evoke dynamic sequences of states in sensory cortical ensembles. Proc. Natl. Acad. Sci. USA 104, 18772–18777 (2007).

    CAS  Article  Google Scholar 

  41. 41.

    Runyan, C. A., Piasini, E., Panzeri, S. & Harvey, C. D. Distinct timescales of population coding across cortex. Nature 548, 92–96 (2017).

    CAS  Article  Google Scholar 

  42. 42.

    Sadacca, B. F. et al. The behavioral relevance of cortical neural ensemble responses emerges suddenly. J. Neurosci. 36, 655–669 (2016).

    CAS  Article  Google Scholar 

  43. 43.

    Churchland, M. M. et al. Stimulus onset quenches neural variability: A widespread cortical phenomenon. Nat. Neurosci. 13, 369–378 (2010).

    CAS  Article  Google Scholar 

  44. 44.

    Liu, H. & Fontanini, A. State dependency of chemosensory coding in the gustatory thalamus (VPMpc) of alert rats. J. Neurosci. 35, 15479–15491 (2015).

    CAS  Article  Google Scholar 

  45. 45.

    Grewe, B. F. et al. Neural ensemble dynamics underlying a long-term associative memory. Nature 543, 670–675 (2017).

    CAS  Article  Google Scholar 

  46. 46.

    Chow, S. S., Romo, R. & Brody, C. D. Context-dependent modulation of functional connectivity: secondary somatosensory cortex to prefrontal cortex connections in two-stimulus-interval discrimination tasks. J. Neurosci. 29, 7238–7245 (2009).

    CAS  Article  Google Scholar 

  47. 47.

    Zenke, F., Agnes, E. J. & Gerstner, W. Diverse synaptic plasticity mechanisms orchestrated to form and retrieve memories in spiking neural networks. Nat. Commun. 6, 6922 (2015).

    CAS  Article  Google Scholar 

  48. 48.

    Litwin-Kumar, A. & Doiron, B. Formation and maintenance of neuronal assemblies through synaptic plasticity. Nat. Commun. 5, 5319 (2014).

    CAS  Article  Google Scholar 

  49. 49.

    Chance, F. S., Abbott, L. F. & Reyes, A. D. Gain modulation from background synaptic input. Neuron 35, 773–782 (2002).

    CAS  Article  Google Scholar 

  50. 50.

    Engel, T. A. et al. Selective modulation of cortical state during spatial attention. Science 354, 1140–1144 (2016).

    CAS  Article  Google Scholar 

  51. 51.

    Zucchini, W. & MacDonald, I.L. Hidden Markov Models for Time Series: An Introduction Using R (CRC Press, 2009).

  52. 52.

    La Camera, G. & Richmond, B. J. Modeling the violation of reward maximization and invariance in reinforcement schedules. PLoS Comput. Biol. 4, e1000131 (2008).

    Article  Google Scholar 

  53. 53.

    Tuckwell, H. C Introduction to Theoretical Neurobiology (Cambridge Univ. Press, 1988).

  54. 54.

    Lánský, P. & Sato, S. The stochastic diffusion models of nerve membrane depolarization and interspike interval generation. J. Peripher. Nerv. Syst. 4, 27–42 (1999).

    PubMed  Google Scholar 

  55. 55.

    Richardson, M. J. Effects of synaptic conductance on the voltage distribution and firing rate of spiking neurons. Phys. Rev. E 69, 051918 (2004).

    Article  Google Scholar 

  56. 56.

    Brunel, N. & Sergi, S. Firing frequency of leaky intergrate-and-fire neurons with synaptic current dynamics. J. Theor. Biol. 195, 87–95 (1998).

    CAS  Article  Google Scholar 

  57. 57.

    Fourcaud, N. & Brunel, N. Dynamics of the firing probability of noisy integrate-and-fire neurons. Neural Comput. 14, 2057–2110 (2002).

    Article  Google Scholar 

  58. 58.

    La Camera, G., Rauch, A., Lüscher, H. R., Senn, W. & Fusi, S. Minimal models of adapted neuronal response to in vivo-like input currents. Neural Comput. 16, 2101–2124 (2004).

    Article  Google Scholar 

  59. 59.

    Rigotti, M. et al. The importance of mixed selectivity in complex cognitive tasks. Nature 497, 585–590 (2013).

    CAS  Article  Google Scholar 

  60. 60.

    Breiman, L. Bagging predictors. Mach. Learn. 24, 123–140 (1996).

    Google Scholar 

Download references


This work was supported by a National Institute of Deafness and Other Communication Disorders Grant no. K25-DC013557 (L.M.), by the Swartz Foundation Award 66438 (L.M.), by National Institute of Deafness and Other Communication Disorders Grant nos. R01DC012543 and R01DC015234 (A.F.), and partly by a National Science Foundation Grant no. IIS­1161852 (G.L.C.). The authors would like to thank S. Fusi, A. Maffei, G. Mongillo, and C. van Vreeswijk for useful discussions.

Author information




L.M., G.L.C., and A.F. designed the project, discussed the models and the data analyses, and wrote the manuscript. L.M. performed the data analysis, model simulations, and theoretical analyses.

Corresponding authors

Correspondence to G. La Camera or A. Fontanini.

Ethics declarations

Competing interests

The authors declare no competing interests.

Additional information

Journal peer review information: Nature Neuroscience thanks Paul Miller and other anonymous reviewer(s) for their contribution to the peer review of this work.

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

Integrated supplementary information

Supplementary Figure 1 Cue responses in data and model.

a: Comparison of cue responses of the clustered network with those published in Ref. 10 (right-most bar). Shown are the peak firing rate responses for different values of cue-induced spatial variance σ2. b: Data from different sensory modalities together with across-modalities average in the rightmost bar from Ref. 10 c: same as a for the homogeneous network with σ2 = 20%. Peak responses (ΔPSTH, mean ± s.e.m. across neurons from n=20 simulated networks, 50 neurons/network were randomly sampled; dots represent single neurons) were computed as the difference between the peak activity post-cue and the average activity pre-cue. Single neuron responsiveness was defined via a change-point analysis (see Methods). In the data, PSTH modulation in cue-responsive neurons was consistent across modalities (excited responses peaked at 12 ± 1.0 spks/s, inhibited responses peaked at −4.8 ± 0.3 spks/s). In the model, cue responses depended only slightly on the cue-induced modulation of the spatial variance σ2 (expressed in percent of baseline; see panel a), and were in quantitative agreement with the data over a wide range of parameters. When σ = 20%, cue responses were: in the clustered network model (see Fig. 1a-c of the main text), 11.1 ± 0.8 spks/s (excited) and −6.2 ± 0.3 spks/s (inhibited); in the homogeneous network (panel c, see Fig. 1d-f of the main text), cue responses had average peaks 11.2 ± 1.2 spks/s (excited) and −3.9 ± 0.2 spks/s (inhibited, significantly different from Data, two-sided t-test, p=0.04). Panels a and c: two-sided t-test with multiple comparison Bonferroni correction, *=p<0.05; n.s.=non-significant. “Data” in panels a and c is the same as “Average” in panel b.

Supplementary Figure 2 Model neurons with no cue response.

Raster plots and PSTH (pin curves) of representative single neuron responses to cue and one stimulus in expected trials in the clustered (a) and homogeneous (b) networks. Top row: non-responsive neurons. Bottom row: neurons that are stimulus-responsive but not cue-responsive. A neuron was deemed responsive if the PSTH was significantly different from baseline (pink horizontal bars: p<0.05, two-sided t-test with multiple-bin Bonferroni correction). The PSTHs report mean ± s.e.m. firing rate across 20 simulated trials with the same stimulus.

Supplementary Figure 3 Classification algorithm used for population decoding.

a: Stimulus decoding is accelerated in the presence of the anticipatory cue. Here we show that similar results are obtained when considering the decoding accuracy for single tastes separately. a: Schematics of the decoding algorithm: The colored curves represent the temporal evolution of neural activity across N simultaneously recorded neurons. The four colors label trajectories obtained with four different stimuli (tastants). Hues help visualize the time course within each trajectory. Each time bin of activity (black dots) was decoded using an independent classifier. Within each condition (‘expected’ or ‘unexpected’), decoding performance was assessed via a cross-validation procedure, yielding a confusion matrix whose diagonal represents the accuracy of classification for each taste in that time bin. b: Time course of decoding accuracy in expected (dashed curves) vs. unexpected (full curves) conditions for each individual tastant (left, color-coded) and for the across-taste average (right). The left panel demonstrates coding anticipation for each tastant separately. Color-coded horizontal bar represents significant difference between decoding accuracy in expected vs. unexpected trials, p<0.05, two-sided t-test with multiple bin Bonferroni correction (notations as in Fig. 1c of the main text).

Supplementary Figure 4 Firing-rate coding of expectation.

Stimulus coding could accelerate if the cue increased the firing rate of the stimulus-selective neurons compared to the non-selective neurons. Here we show that coding anticipation in the clustered network is not driven by such cue-induced changes in firing rate selectivity. To prove this point, we estimated the time course of the firing rate difference Δr between stimulus-selective and nonselective neurons in the expected (pink) and unexpected (blue) conditions in the clustered and homogeneous networks. We found no difference in Δr between conditions in the clustered network (a), demonstrating that anticipatory activity is not driven by changes in firing rates. In the homogeneous network (b), Δr was larger in the unexpected condition, in agreement with the reversed trend in coding speed found in Fig. 1f. Panel b: black horizontal bar: p<0.05, two-sided t-test with multiple bin Bonferroni correction. Both panels: main curves represent means of Δr over 20 sessions; shaded area represents s.e.m.

Supplementary Figure 5 Robustness of anticipatory activity.

a: Anticipatory activity is present in the case of step-like stimuli (top: time course of stimulus as a fraction of baseline afferent current; bottom: time course of decoding accuracy; notations as in Fig. 1c of main text). Inset: aggregate analysis across n=20 simulated networks of the onset times of significant decoding (mean ± s.e.m.) in expected (pink) vs. unexpected trials (blue) shows significantly faster onsets in the expected condition (two-sided t-test, p=4.0x10-4). b: Anticipatory activity did not depend on the number of stimuli presented to the network. Latency of significant stimulus decoding was faster in expected (pink) compared to unexpected (blue) trials with up to 16 stimuli (clusters were selective to a given stimulus with 50% probability; error bars represent mean ± s.e.m. across n=20 simulated networks). c: Anticipatory activity was present in the case of step-like cue with spatial variance σ=20% and a linearly ramping stimulus as in main Fig. 2a (top: time course of stimulus as a fraction of baseline afferent current; inset notations as in panel a: two-sided t-test, p=1.5x10-4). d: Anticipatory activity was present even with overlapping clusters33. In this model, neurons had a probability f=0.06 of belonging to one of the Q=14 clusters, with a fraction of \(f^k\left( {1 - f} \right)^{Q - k}\) neurons belonging to any set of k specific clusters and E-to-E synaptic connections given by \(J_{ij} = p_{EE}({\it{\epsilon }}_{ij}\xi _ + + \left( {1 - {\it{\epsilon }}_{ij}} \right)\xi _ - )\). Here, pEE=0.2 is the connection probability; ξ± are the synaptic weights values sampled from normal distributions with means J± and variances \(\delta ^2J_ \pm ^2\), respectively (δ=0.01). The synaptic weights were potentiated with probability \({\it{\epsilon }}_{ij} = \frac{{P_{ij}}}{{P_{ij} + \rho \,fD_{ij}}}\), where \(P_{ij} = \Sigma _{k = 1}^Q\eta _i^k\eta _j^k\) is the number of clusters in common between neurons i and j (\(\eta _i^k = 1\) if cluster k contains neuron i and \(\eta _i^k = 0\) otherwise), while \(D_{ij} = \Sigma _{k = 1}^Q\eta _i^k(1 - \eta _j^k)\), with ρ=2.75 (see Ref. 30 for more details on this model). The E-to-I, I-to-E, and I-to-I connection probability, the stimuli and the anticipatory cue were the same as for the clustered networks (Table 1 in the main text). The remaining parameters of the network are reported in Supplementary Table 1 (inset notations as in panel a: two-sided t-test, p=1.1x10-3). Main panels: *=p<0.05, **=p<0.01, ***=p<0.001, post-hoc t-test with Bonferroni correction. Horizontal black bar, p<0.05, two-sided t-test with multiple-bin Bonferroni correction. Insets: **=p<0.01, ***=p<0.001, two-sided t-test.

Supplementary Figure 6 A distracting cue slows down stimulus coding (model).

To show that the anticipatory effect of our model cue is specific, we give here an example of a manipulation leading to the opposite effect. Specifically, if the cue is modeled as an increase in the mean input current to the inhibitory population (we refer to such a cue as a “distractor”), stimulus decoding is slowed down rather than accelerated. As we show in the following panels, the coding delay following such “distracting cue” is the consequence of increased energy barriers (panel d) causing slower transition dynamics. In turn, this is due to the sharpening of the effective transfer functions of the excitatory neurons, reflecting increased stability induced by the increased inhibition. a: Schematics of clustered network architecture and stimulation (notations as in Fig. 1a of the main text, here the cue targets the inhibitory neurons). b: Time course of cross-validated decoding accuracy during distracted (brown) trials was slower than during unexpected (red) trials (notations as in Fig. 1c). Inset: aggregate analysis across n=20 simulated networks, mean ± s.e.m.; two-sided t-test, p=1.3x10-3). c: Activation latency of stimulus-selective clusters after stimulus presentation was delayed during distracted trials (mean ± s.e.m. across n=20 simulated networks, two-sided t-test, p=1.5x10-13). d: Mean field theory of simplified 2-cluster network (notations and model as in Fig. 4c, lighter brown denotes stronger stimuli). Left panel: the transition probability from the non-coding (right well) to the coding state (left well) increased with larger stimuli. In ‘distracted trials’ (dashed curves) the barrier height Δ from the non-coding to the coding state is larger than in unexpected trials (full curves), leading to slower coding in the distracted condition. Right panel: effective energy barriers as a function of stimulus intensity, with (full lines) and without the cue (dashed). Panels b and c: **=p<0.01, ***=p<0.001, two-sided t-test.

Supplementary Figure 7 Specificity of anticipatory activity (part 1).

We compared our spatial variance model (a, see Fig. 1a-c and Supplementary Fig. 1a for notations) to alternative models where the anticipatory cue modulates the feedforward couplings Jext (b) or the recurrent couplings JEE (c). All models had the same architecture and cue temporal profile of the main clustered network model (top row panels). In the alternative models, the cue targeted all clustered E neurons. Models were scored on the ability to match the experimental data on typology of cue response and amount of stimulus-coding anticipation. Cue responses were quantified as the ΔPSTH = peak cue response – baseline firing rate as in Supplementary Fig. 1 (bottom left panels; green: excited neurons, red: inhibited neurons; ΔPSTH, mean ± s.e.m. across neurons from n=10 simulated networks, 50 neurons/network were randomly sampled; dots represent single neurons), while coding anticipation was assessed via latency of stimulus decoding (bottom right panels, notations as in inset of Fig. 1c; aggregate analysis across n=10 simulated networks of the onset times of significant decoding (mean ± s.e.m.) in expected (pink) vs. unexpected trials (blue)). In the feedforward coupling model (b), the cue was a time-dependent modulation of the external synaptic coupling Jext = JE0, identical for all clustered excitatory neurons. Cue responses were heterogeneous but significantly different from the experimental data (bottom left, two-sided t-test; excited responses: 10% with p=6.3x10-12, 20% with p=4.0x10-3, inhibited responses: 10% non-significant, 20% with p=0.01); coding anticipation was absent for either moderate or strong positive modulations (10%-20% above baseline; bottom right). For negative cue modulations, only inhibited cue responses were observed and no coding anticipation was present (not shown). In the recurrent coupling model (c), the cue was a time-dependent modulation of the E-to-E recurrent coupling strength JEE. For negative JEE modulation (the more likely to produce a faster dynamics), coding anticipation was present (bottom right, two-sided t-test; 10% with p=0.0039, 20% with p=0.004), however, peak cue responses were strongly inhibited over a wide range of parameters, thus incompatible with the empirical data (bottom left, two-sided t-test; excited responses: 10% with p=0.01, 20% with p=0.0097, inhibited responses: 10% with p=6.9x10-6, 20% with p=2.9x10-6). We obtained similar results after decreasing the spike thresholds of the excitatory neurons (down to -30%, not shown). Increasing the thresholds of inhibitory neurons had no significant impact on the behavior of the model (up to 30%, not shown). For positive JEE modulation (not shown), coding anticipation was absent and cue responses were mostly excited. Non-modulated network parameters were as in Table 1 (all panels). All panels: two-sided t-test, *=p<0.05, **=p<0.01, ***=p<0.001.

Supplementary Figure 8 Specificity of anticipatory activity (part 2).

We compared our model (a, see Fig. 1a-c and Supplementary Fig. 1a) to an alternative model where the anticipatory cue modulates the background synaptic input, driving a simultaneous increase in background noise and shunting inhibition as in Ref. 59 (b-c). The baseline noise level was modeled after an Ornstein-Uhlenbeck process with zero mean and variance σext = 0.5rext, where rext was the mean afferent current. The cue increased the background noise by a factor X: \(\sigma _{ext}^2 \to X\sigma _{ext}^2\) while shunting the membrane time constant by a factor 1/X: τmτm / X (we refer to Ref. 59 for details). Both modulations followed the same double exponential time course of the spatial variance model of the main text (top panels in b-c). b: With even a moderate factor of X=1.5, the cue induced mostly inhibited cue responses (bottom left panels; green: excited neurons, red: inhibited neurons; ΔPSTH, mean ± s.e.m. across neurons from n=10 simulated networks, 50 neurons/network were randomly sampled; dots represent single neurons), significantly different from the experimental data (two-sided t-test; excited responses: n/a; inhibited responses: p=1.9x10-5) and no anticipation (bottom right, aggregate analysis across n=10 simulated networks of the onset times of significant decoding (mean ± s.e.m.) in expected (pink) vs. unexpected trials (blue); two-sided t-test, non significant) due to the strong shunting effect on the network dynamics, leading to a strong reduction of excitability in the clusters. At X=2, excitatory neurons were transiently silenced (not shown). c: In an effort to obtain a fair comparison with the spatial variance model, we reduced the shunting effect (using peak value τm / Xϵ, with X=1.5 and ϵ=1/4) while keeping the same amount of background noise. In this case, cue responses were more similar to the data (two-sided t-test; excited responses: p=1.5x10-7; inhibited responses: n/a), but coding slowed down compared to the unexpected condition (bottom right, two-sided t-test, p=0.015). All panels: two-sided t-test, *=p<0.05, **=p<0.01, ***=p<0.001.

Supplementary Figure 9 Specificity of anticipatory activity (part 3).

In the alternative model where the cue modulates the background synaptic input (Supplementary Fig. 8), we explored the parameter space by independently scaling background noise, shunting, and mean afferent current rext (transiently increased with the same time course as the other quantities to counteract the shunting effect, see top panels in Supplementary Fig. 8b-c), but found that coding anticipation was never present (bottom row). We concluded that a model cue inducing an increase in background synaptic activity did not lead to anticipatory activity. a: ΔPSTH for X=1.5 (full bar: excited neurons; dashed bar: inhibited neurons; ΔPSTH, mean ± s.e.m. across neurons from n=10 simulated networks, 50 neurons/network were randomly sampled; dots represent neurons with excited (green) and inhibited (red) responses); b: decoding latency (aggregate analysis across n=10 simulated networks of the onset times of significant decoding (mean ± s.e.m.) in expected trials; dashed blue line: coding latency in unexpected trials, mean ± s.e.m.) as a function of the scaling parameter X for the 5 scaling regimes shown in a (same color code). Color code: Black, \(\sigma _{ext}^2 \to X\sigma _{ext}^2,r_{ext} \to r_{ext},\tau _m \to \tau _m{\mathrm{/}}X\) (same as in Supplementary Fig. 8b); Dark grey, \(\sigma _{ext}^2 \to X\sigma _{ext}^2,r_{ext} \to X^{\frac{1}{2}}r_{ext},\tau _m \to \tau _m{\mathrm{/}}X\); Light grey, \(\sigma _{ext}^2 \to X\sigma _{ext}^2,r_{ext} \to Xr_{ext},\tau _m \to \tau _m{\mathrm{/}}X\); Dark brown: \(\sigma _{ext}^2 \to X\sigma _{ext}^2,r_{ext} \to r_{ext},\tau _m \to \tau _m{\mathrm{/}}X^{\frac{1}{2}}\); Light brown: \(\sigma _{ext}^2 \to X\sigma _{ext}^2,r_{ext} \to r_{ext},\tau _m \to \tau _m{\mathrm{/}}X^{\frac{1}{4}}\) (same as in Supplementary Fig. 8c). Panel a, two-sided t-test: *=p<0.05, **=p<0.01, ***=p<0.001.

Supplementary information

Supplementary Figures 1–9

Supplementary Figures 1–9 and Supplementary Tables 1 and 2.

Reporting Summary

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Mazzucato, L., La Camera, G. & Fontanini, A. Expectation-induced modulation of metastable activity underlies faster coding of sensory stimuli. Nat Neurosci 22, 787–796 (2019). https://doi.org/10.1038/s41593-019-0364-9

Download citation

Further reading