Trends in Cognitive Sciences
Volume 18, Issue 9, September 2014, Pages 480-487
Journal home page for Trends in Cognitive Sciences

Review
Scale-free brain activity: past, present, and future

https://doi.org/10.1016/j.tics.2014.04.003Get rights and content

Highlights

  • Arrhythmic, scale-free brain activity is distinct from brain oscillations.

  • Scale-free brain activity contains rich temporal structures beyond power spectrum.

  • Scale-free brain activity is relevant to task performance and arousal state.

  • Scale-free brain activity is altered in developmental and disease processes.

  • Computational modeling has shed light on the potential generative mechanisms of scale-free brain activity.

Brain activity observed at many spatiotemporal scales exhibits a 1/f-like power spectrum, including neuronal membrane potentials, neural field potentials, noninvasive electroencephalography (EEG), magnetoencephalography (MEG), and functional magnetic resonance imaging (fMRI) signals. A 1/f-like power spectrum is indicative of arrhythmic brain activity that does not contain a predominant temporal scale (hence, ‘scale-free’). This characteristic of scale-free brain activity distinguishes it from brain oscillations. Although scale-free brain activity and brain oscillations coexist, our understanding of the former remains limited. Recent research has shed light on the spatiotemporal organization, functional significance, and potential generative mechanisms of scale-free brain activity, as well as its developmental and clinical relevance. A deeper understanding of this prevalent brain signal should provide new insights into, and analytical tools for, cognitive neuroscience.

Introduction

A student entering neuroscience today might learn about the irregular, Poisson-like firing in cortical pyramidal neurons on the one hand, and the plethora of brain oscillations on the other hand. Both are well-established neuroscience phenomena: the former from single- or multiunit recordings of neuronal spiking, the latter from recordings of brain electrical field potentials, such as local field potentials (LFP), EEG, and MEG. Why is it that one modality has emphasized irregular patterns of neural activity, whereas the other has emphasized oscillatory patterns? In fact, regular, rhythmic neuronal firing patterns do exist in cortical excitatory neurons; they are just less common 1, 2, 3 (Figure 1A). Irregular, arrhythmic (see Glossary) field potential activity patterns also exist (Figure 1B) and account for the majority of the signal power recorded in LFP, EEG, and/or MEG experiments (Figure 1C), but are less studied than brain oscillations. In this review, I focus on what we currently know about this prevalent, arrhythmic component of brain field potentials, and identify several urgent questions in this research field.

Section snippets

Brain oscillations versus scale-free brain activity

Brain oscillations are recurring patterns of brain activity that follow a particular temporal beat. For example, the first discovered EEG rhythm, the occipital alpha wave, proceeds at approximately 10 cycles per second [4]. Thus, brain oscillations are most easily identified in the frequency domain, because their power spectra contain peaks at the corresponding frequency ranges (arrows in Figure 1C). There are several brain oscillations at different frequency ranges, each with their own

Scale-free brain activity is not unstructured noise

1/f-type temporal dynamics are prevalent not only in the nervous system, but also in nature at large 36, 37. The ubiquity of scale-free dynamics in a variety of systems was often taken as evidence that these dynamics lack functional specificity, as exemplified by the colloquial name ‘1/f noise.’ To a large extent, the historical neglect of scale-free brain activity is due to this deflationary interpretation of scale-free dynamics. However, it is important to keep in mind that the power spectrum

Beyond nested frequencies

Nonetheless, what do these nested-frequency patterns in scale-free dynamics mean? Nested-frequency analysis necessitates filtering the broadband signal in different frequency ranges and extracting the phase and power of a lower- and a higher- frequency band, respectively. Whereas it is straightforward to characterize the phase and power corresponding to rhythmic brain oscillations, the interpretation of phase and power extracted from filtered arrhythmic signals requires more caution. Although

Functional significance of scale-free brain activity

Research on the functional roles of scale-free brain activity is just beginning. Nonetheless, there are several tantalizing lines of evidence suggesting that it is intimately related to brain functioning.

First, the broadband (approximately 5∼200 Hz) power of LFPs has been shown to correlate tightly with population neuronal firing rates in both human and macaque 14, 46. Ray and Maunsell [46] presented an impressive dissociation between scale-free brain activity and brain oscillation in the same

Generative mechanisms of scale-free brain activity

What are the generative mechanisms of scale-free brain activity? In this section, I first focus on existing computational modeling studies on the power spectral shape of invasively recorded LFP and ECoG signals, and then describe the discrepancy of the power-law exponent observed across modalities and the intriguing questions posed by these findings.

Relation between broadband activity, amplitude fluctuations of brain oscillations, and neuronal avalanches

In this review, I focused on scale-free dynamics in the raw fluctuations of broadband (from <0.01 Hz to ∼500 Hz) electrical and magnetic signals from the brain, as well as the low-frequency (<0.5 Hz) activity recorded in fMRI. As mentioned above, the amplitude fluctuations of narrow-band brain oscillations also exhibit scale-free dynamics [20]; moreover, a recent burgeoning literature has demonstrated the functional 26, 27, 59, developmental [29], and clinical [34] relevance of this phenomenon.

Acknowledgments

This work was supported by the Intramural Research program of the NIH/NINDS. I thank Xiao-Jing Wang and Rishidev Chaudhuri for discussions; Qi Li and Zachary Hill for collecting resting-state MEG and EEG data for estimating the power-law exponent; and Brian Maniscalco and three anonymous reviewers for helpful feedback on previous drafts.

Glossary

Arrhythmic
without a predominant temporal frequency, often used interchangeably with ‘aperiodic’ and ‘irregular’.
Brain oscillations
brain activity patterns that recur with a particular temporal frequency.
Exponentially decaying function
dXdt=λX, where X is a time-varying function and λ is the decay rate. It is called ‘exponentially decaying’ because the solution to this function is: X(t) = X0eλt. Convolution of two exponential functions can be written as: dVdt=λ1V+I and dIdt=λ2I+ε. In the context

References (87)

  • M. Wei

    Identifying major depressive disorder using Hurst exponent of resting–state brain networks

    Psychiatry Res.

    (2013)
  • B.J. He et al.

    The fMRI signal, slow cortical potential and consciousness

    Trends Cogn. Sci.

    (2009)
  • P. Fries

    Finding gamma

    Neuron

    (2008)
  • C.J. Honey

    Slow cortical dynamics and the accumulation of information over long timescales

    Neuron

    (2012)
  • W.J. Pan

    Infraslow LFP correlates to resting-state fMRI BOLD signals

    Neuroimage

    (2013)
  • I. Kahn

    Optogenetic drive of neocortical pyramidal neurons generates fMRI signals that are correlated with spiking activity

    Brain Res.

    (2013)
  • C.D. Brody

    Basic mechanisms for graded persistent activity: discrete attractors, continuous attractors, and dynamic representations

    Curr. Opin. Neurobiol.

    (2003)
  • G. Major et al.

    Persistent neural activity: prevalence and mechanisms

    Curr. Opin. Neurobiol.

    (2004)
  • W.J. Freeman

    Spatial spectra of scalp EEG and EMG from awake humans

    Clin. Neurophysiol.

    (2003)
  • N.K. Logothetis

    In vivo measurement of cortical impedance spectrum in monkeys: implications for signal propagation

    Neuron

    (2007)
  • R.T. Canolty et al.

    The functional role of cross-frequency coupling

    Trends Cogn. Sci.

    (2010)
  • D. Plenz et al.

    The organizing principles of neuronal avalanches: cell assemblies in the cortex?

    Trends Neurosci.

    (2007)
  • W.R. Softky et al.

    The highly irregular firing of cortical cells is inconsistent with temporal integration of random EPSPs

    J. Neurosci.

    (1993)
  • W. Bair

    Power spectrum analysis of bursting cells in area MT in the behaving monkey

    J. Neurosci.

    (1994)
  • H. Berger

    Uber das Elektroenkephalogramm des Menschen

    Arch. Psyhiatr. Nervenkr.

    (1929)
  • P. Fries

    Neuronal gamma-band synchronization as a fundamental process in cortical computation

    Annu. Rev. Neurosci.

    (2009)
  • A. Destexhe

    The high-conductance state of neocortical neurons in vivo

    Nat. Rev. Neurosci.

    (2003)
  • S. El Boustani

    Network-state modulation of power-law frequency-scaling in visual cortical neurons

    PLoS Comput. Biol.

    (2009)
  • J. Milstein

    Neuronal shot noise and Brownian 1/f2 behavior in the local field potential

    PLoS ONE

    (2009)
  • K.J. Miller

    Power-law scaling in the brain surface electric potential

    PLoS Comput. Biol.

    (2009)
  • J.R. Manning

    Broadband shifts in local field potential power spectra are correlated with single-neuron spiking in humans

    J. Neurosci.

    (2009)
  • W.J. Freeman et al.

    Simulated power spectral density (PSD) of background electrocorticogram (ECoG)

    Cogn. Neurodyn.

    (2009)
  • N. Dehghani

    Comparative power spectral analysis of simultaneous electroencephalographic and magnetoencephalographic recordings in humans suggests non-resistive extracellular media: EEG and MEG power spectra

    J. Comput. Neurosci.

    (2010)
  • B.J. He

    Scale-free properties of the functional magnetic resonance imaging signal during rest and task

    J. Neurosci.

    (2011)
  • E. Bullmore

    Colored noise and computational inference in neurophysiological (fMRI) time series analysis: resampling methods in time and wavelet domains

    Hum. Brain Mapp.

    (2001)
  • P. Ciuciu

    Scale-free and multifractal time dynamics of fMRI signals during rest and task

    Front. Physiol.

    (2012)
  • K. Linkenkaer-Hansen

    Long-range temporal correlations and scaling behavior in human brain oscillations

    J. Neurosci.

    (2001)
  • S.B. Lowen

    Quantal neurotransmitter secretion rate exhibits fractal behavior

    J. Neurosci.

    (1997)
  • D.L. Gilden

    Cognitive emissions of 1/f noise

    Psychol. Rev.

    (2001)
  • D.J. Smit

    Long-range temporal correlations in resting-state alpha oscillations predict human timing-error dynamics

    J. Neurosci.

    (2013)
  • J.M. Palva

    Neuronal long-range temporal correlations and avalanche dynamics are correlated with behavioral scaling laws

    Proc. Natl. Acad. Sci. U.S.A.

    (2013)
  • S. Monto

    Very slow EEG fluctuations predict the dynamics of stimulus detection and oscillation amplitudes in humans

    J. Neurosci.

    (2008)
  • D.J. Smit

    Scale-free modulation of resting-state neuronal oscillations reflects prolonged brain maturation in humans

    J. Neurosci.

    (2011)
  • Cited by (454)

    View all citing articles on Scopus
    View full text