Five methodological challenges in cognitive electrophysiology

1

2Q5

3Q1

45

6

78910111213141516171819

31

32

3334

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

NeuroImage xxx (2013) xxx–xxx

YNIMG-10747; No. of pages: 9; 4C: 4

Contents lists available at ScienceDirect

NeuroImage

j ourna l homepage: www.e lsev ie r .com/ locate /yn img

Review

Five methodological challenges in cognitive electrophysiology

F

Michael X. Cohen a,⁎, Rasa Gulbinaite b

a Department of Psychology, University of Amsterdam, The Netherlandsb Department of Psychology, University of Groningen, The Netherlands

⁎ Corresponding author.E-mail address: [email protected] (M.X. Cohen)

1053-8119/$ – see front matter © 2013 Elsevier Inc. All rihttp://dx.doi.org/10.1016/j.neuroimage.2013.08.010

Please cite this article as: Cohen, M.X., Gulbindx.doi.org/10.1016/j.neuroimage.2013.08.01

O

a b s t r a c t
a r t i c l e i n f o
20

21

22

23

24

25

26

27

28

Article history:Accepted 6 August 2013Available online xxxx

Keywords:OscillationsCognitionElectrophysiologyTime–frequency analysesCausality

PROHerewe discuss fivemethodological challenges facing the current cognitive electrophysiology literature that ad-

dress the roles of brain oscillations in cognition. The challenges focus on (1) unambiguous and consistent termi-nology, (2) neurophysiologically meaningful interpretations of results, (3) evaluation and comparison ofdifferent spatial filters often used inM/EEG research, (4) the role of multiscale interactions in brain and cognitivefunction, and (5) development of biophysically plausible cognitive models. We also suggest research directionsthat will help address these challenges. We hope that this paper will help foster discussions and debates aboutimportant themes in the study of how the brain's rhythmic patterns of spatiotemporal electrophysiological activ-ity support cognition.

© 2013 Elsevier Inc. All rights reserved.

2930
D
Contents

RECTE

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0Challenge 1: Widespread agreement on analysis terminology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0Challenge 2: Neurophysiological interpretation of time–frequency results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0Challenge 3: Reconciling the many diverse spatial filters used in cognitive electrophysiology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0Challenge 4: Characterizing multi-scale interactions in neuroelectric activity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0Challenge 5: Developing neurophysiologically grounded psychological theories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0New horizons in cognitive electrophysiology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0

60

61

62

63

64

65

66

67

68

69

70

71

72

73

74

UNCO

RIntroduction

Cognitive electrophysiology is a field that bridges neuroscience andpsychology, and focuses on understanding how cognitive functions (in-cluding perception, memory, language, emotions, behavior control, andsocial cognition) are supported or implemented by the electrical activityproduced by populations of neurons. The main methodological toolsused by cognitive electrophysiologists are EEG andMEG, and intracrani-al recordings such as electrocorticogram and single- and multi-unit re-cordings. Although these methods span a range of species and spatialscales, they all share the common feature that they measure electro-magnetic activity. Thus, the major assumption underlying the broadspectrum of cognitive electrophysiology studies is that one key neuralmechanism of processing and transferring information is (or, at least,can be understood through) electrical activity.

The purpose of this paper is to highlight and discuss five majormethodological challenges facing cognitive electrophysiology. Some of

75

76

77.

ghts reserved.

aite, R., Five methodological0

these challenges are related to each other; discussing them individuallyis done mainly for convenience. Indeed, in several cases, addressingone challenge may help address other challenges. We focus mainlyon methods and data analyses involving time–frequency-based ap-proaches, because these are the most rapidly developing methodologi-cal approaches in cognitive electrophysiology, and, as will be describedbelow, have a large potential for understanding neurophysiological pro-cesses underlying cognitive operations.

Some readers may disagree with the importance of some of thesechallenges, or could name additional challenges than the five presentedhere. Nonetheless, we hope that this paper will help catalyze furtherdiscussions in current trends and important future directions in cogni-tive electrophysiology.

Challenge 1: Widespread agreement on analysis terminology

Consider the following statistical analysis terms: correlation, ANOVA,factor analysis, and receiver operating characteristic (ROC). Whensomeone says that they performed an ANOVA, there is no ambiguityabout which sets of equations were applied to the data. Furthermore,

challenges in cognitive electrophysiology, NeuroImage (2013), http://

http://dx.doi.org/10.1016/j.neuroimage.2013.08.010

mailto:[email protected]


http://www.sciencedirect.com/science/journal/10538119



TD

78

79

80

81

82

83

84

85

86

87

88

89

90

91

92

93

94

95

96

97

98

99

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

120

121

122

123

124

125

126

127

128

129

130

131

132

133

134

135

136

137

138

139

140

141

142

143

144

145

146

147

148

149

150

151

152

153

154

155

156

157

158

159

160

161

162

163

164

165

166

167

168

169

170

171Q2

172

173

174

175

176

177

178

179

180

t1:1

t1:2

t1:3

t1:4

t1:5

t1:6

t1:7

t1:8

2 M.X. Cohen, R. Gulbinaite / NeuroImage xxx (2013) xxx–xxx

NCO

RREC

even though the term ROC provides little insight on the mathematicalprocedure underlying that analysis, most people with a background inengineering, math, psychology, or physics will knowwhat an ROC anal-ysis implies and how the results can be interpreted.

Precision and widespread agreement in analysis terminology islacking in cognitive electrophysiology. This is problematic because in-consistent, ambiguous, or confusing terminology impedes cross-studycomparisons and theory development (Gardiner and Java, 1993;Tulving, 2000). To illustrate this point, consider the following electro-physiological data analysis terms: synchronization, event-related spec-tral perturbation, time–frequency response, and connectivity. Theseand other terms are ambiguous and are often lab- or software-specific. When someone says that they found an increase in alpha syn-chronization, youdonot knowwhether theymean an increase in powerat one electrode or an increase in phase-based connectivity betweentwo electrodes. This confusion arises because some researchers usethe term “synchronization” to indicate the squared amplitude of the fre-quency band-specific filtered signal at one electrode (Pfurtscheller,1992), whereas other researchers use this same term to indicate consis-tency in phase angle differences between two electrodes. However,these two analyses have very different interpretations, putative neuro-physiological origins, theoretical implications, and methodologicalconcerns. Terms like spectral perturbation (Makeig, 1993) and time–frequency response are also ambiguous, because they could refer tospectral changes expressed in power, phase, connectivity, band-specific network properties, or any number of other features of time–frequency-based analyses.

Within a field of science, there should be a one-to-one mappingbetween terms and their meanings (also called the incontrovertibilityof terms rule; Gardiner and Java, 1993). However, cognitive electrophys-iology suffers from a many-to-many terminology mapping problem:the same term can have different meanings (e.g., the term “synchroni-zation,” as described in the previous paragraph); and different termscan indicate the same mathematical procedure (e.g., inter-trial phasecoherence vs. phase-locking index/value can refer to the same analysis,which assesses the consistency of phase angles at one electrode-time–frequency point over trials). The many-to-many mapping of analysisterms to mathematical procedures slows scientific progress by creatingconfusion about how to interpret findings reported in results sections,and how to compare results across studies that use different terms.

Another confusing and ill-defined—but often used—term is “activa-tion.” A brain region is said to be activated (or deactivated) if its activityincreases (or decreases) with respect to a baseline or control condition.Although this term is widely used in univariate fMRI analyses and rela-tively simple analyses of action potential data such as average spike rate,this term becomes less tractable for multi-dimensional electrophysio-logical activity such as field potentials (Singh, 2012). For example, ifa brain region exhibits an increase in inter-trial phase clustering in thetheta band, a decrease in alpha-band power, no change in gamma-band power, and an increase in theta–gamma coupling, is this brain re-gion activated or deactivated? In some cases, increases in power thatseem to lack a clear frequency structure are referred to as “activation”(Burke et al., 2013; Miller et al., 2009), but this approach may obscure
U
Table 1Suggested terminology for time-–frequency-based M/EEG data analyses. See Cohen (2014), fo

Preferred term Description

Power Squared amplitude of frequency-band specific time serieInter-trial-phase-clustering Length of average vector from a distribution of unit phas

time–frequency point over trials.Inter-site-phase-clustering Length of average vector from a distribution of unit phas

between two electrodes at one time–frequency point ove

Notes. Terms are less preferred if they are ambiguous, imprecise, or are interpretations of pusynchronization; ERD = event-related desynchornization; ERSP = event-related spectral pert

Please cite this article as: Cohen, M.X., Gulbinaite, R., Five methodologicaldx.doi.org/10.1016/j.neuroimage.2013.08.010

PRO

OF

the fine temporal structure of activity, such asmultiple overlapping fre-quencies (Crone et al., 2011) or temporal or correlation-based informa-tion coding (Engel et al., 1992; Tsukada et al., 1996).

Perhaps the lack of terminological convergencewas less of a concerna few decades ago, when few research groups were performing time–frequency-based analyses, andmost analyseswere based on the squaredamplitude of the frequency band-specific signal (i.e., power). However,the lack of consistency in analysis terminology becomes problematicas more scientists begin applying sophisticated data analyses. With var-ied and sometimes ambiguous terminology, rapid and efficient cross-study comparisons become increasingly difficult.

The challenge, therefore, is to adopt a widely accepted andunambiguous terminology for describing multivariate changes inelectrophysiological data. We recommend using analysis terms thatclosely and succinctly reflect the mathematical procedure applied tothe data (Cohen, 2014), rather than using terms that reflect interpreta-tions of putative neurophysiological events underlying time–frequencyfeatures. For example, when extracting the energy of a frequency band-specific signal (the squared amplitude), the term “power” should bepreferred over terms such as “synchronization” because “power” is anunambiguous description of the analysis, whereas synchronization is aspeculative interpretation of a result (in this case, that the neural net-works measured by the electrode became synchronized; Pfurtschellerand Lopes da Silva, 1999). At least in the context of electrophysiologydata, it might be best simply to avoid using functional univariateterms like “de/activation.” Instead, terms could describe the statisticalproperties of the data, such as “relatively increased power in the beta-band,” or “correlation between alpha phase and gamma power.” InTable 1, we suggest analysis terms for some commonly used analyses.

EChallenge 2: Neurophysiological interpretation oftime–frequency results

Themathematical development of time–frequency-based data anal-yses, and their applications to studying cognition, has advanced beyondthe understanding of the neurophysiological events that might underliethe results of those analyses. For example, the difference betweenphase-locked and non-phase-locked (also known as evoked and in-duced, respectively) activities remains unclear, with theory and modelssuggesting complex interactions between neurobiological eventsthat may be measured as phase-locked vs. non-phase-locked events(Burgess, 2012; David et al., 2006; McLelland and Paulsen, 2009;Tallon-Baudry and Bertrand, 1999), but little empirical data to providefirm conclusions. Another example is functional connectivity estimatedbetween two electrodes, which can be based on correlations in frequen-cy band-specific power time series (Bruns et al., 2000), or on a cluster-ing of phase value differences (Lachaux et al., 1999). It is unclearwhether connectivity based on power and on phase reflects similarmechanisms (e.g., long-range activation of inhibitory interneurons;Bush and Sejnowski, 1996), and it is unknownwhether the samemech-anisms underlie connectivity in different frequency bands or in differentbrain regions.

r more in-depth discussions and justifications of each term.

Examples of less preferred terms

s Synchronization/desynchronization, ERS/ERD, ERSP, TFRe angles at one Phase-locking, phase-coherence, phase-reset

e angle differencesr trials.

Phase-locking, phase coherence, coherence, synchronization,coupling, phase correlation

tative neural events rather than descriptions of analysis methods. ERS = event-relatedurbation; TFR = time–frequency response.




T

181

182

183

184

185

186

187

188

189

190

191

192

193

194

195

196

197

198

199

200

201

202

203

204

205

206

207

208

209

210

211

212

213

214

215

216

217

218

219

220

221

222

223

224

225

226

227

228

229

230

231

232

233

234

235

236

237

238

239

240

241

242

243

244

245

246

247

248

249

250

251

252

253

254

255

256

257

258

259

260

261

262

263

264

265

266

267

268

269

270

271

272

273

274

275

276

277

278

279

280

281

282

283

284

285

286

287

288

289

290

291

292

293

294

295

296

297

298

299

300

301

302

303

304

305

306

307

308

309

310

3M.X. Cohen, R. Gulbinaite / NeuroImage xxx (2013) xxx–xxx

UNCO

RREC

The challenge, therefore, is to gain a better understanding of theneurophysiological events that lead to time–frequency phenomena ob-served at the scalp, including power, phase, and various measures ofconnectivity. Of course,much is known about the biophysical propertiesof neurons and volume conduction through head tissues that allow theM/EEG to be measured (Lopes da Silva, F. and van Rotterdam, 1982).However, much less is known about the kinds of cellular, synaptic, neu-rochemical, and system-level processes that produce the complexmulti-frequency dynamics that have been linked to cognition. Meetingthis challenge will move cognitive electrophysiology forward: Atpresent, EEG data are typically treated and discussed in terms of ab-stract “time-varying brain signals.” Although this is an accurate descrip-tion, the patterns observed in EEG data are not arbitrary signals, butrather, are direct measurements of complex biophysical processesfrom which cognition, emotion, language, and myriad other func-tions emerge. As patterns in EEG data can be better linked to bothmicro- and mesoscopic-level neural processes, the understanding ofneural machinery and computations underlying cognitive processeswill improve.

This challenge has already received some empirical and theoreticalattention (Wang, 2010). For example, gamma oscillations are knownto be driven by interactions between excitatory and inhibitory cells(Buzsáki andWang, 2012). Other features of EEG data have less clear or-igins. For example, response conflict elicits increased activity in thetheta-band (4–8 Hz) over midfrontal regions (Cavanagh et al., 2009;Cohen, 2011; van Steenbergen et al., 2012).Why does response conflictelicit midfrontal theta, and not alpha, or beta, or gamma? What doesthis indicate about the neural computations that identify and resolveconflict? At present, there are few clear answers to this and manyother questions about why specific cognitive processes are associatedwith specific patterns of spectral responses (Donner and Siegel, 2011).

Not all relevant EEG dynamics are localized to specific frequencybands. Indeed, some task-related electrophysiological features seem tolack frequency band specificity, instead comprising broadband changesin power (Burke et al., 2013; Crone et al., 2011; He et al., 2010;Manning et al., 2009;Miller et al., 2009). This broadbandactivitymay re-flect increases in asynchronous multi-unit activity, or frequency band-specific but transient patterns (Crone et al., 2011; Rieke, 1999). Indeed,broadband gamma activity (60–200 Hz) seems to correlate morestrongly with inter-neuronal synchrony than with average firing rate(Ray et al., 2008).

Of thefive challenges discussed here, this one is themost difficult forcognitive electrophysiologists who study humans to address. This is be-cause understanding the neurophysiological bases of non-invasive scalpM/EEG signals requires methodological approaches that are outside thetypical cognitive electrophysiology lab and expertise, such as simulta-neous invasive and non-invasive recordings. Some of this type of re-search has already been done, for example by studying the relationshipbetween individual neurons and field potentials (Coenen, 1995; Krautet al., 1985; Schroeder et al., 1991; Whittingstall and Logothetis, 2009).Other studies have provided evidence that local field potentials aredriven mainly by synchronous spiking or bursting of cell assemblies,whereas individual spikes (which may reflect neural noise) contributesignificantly less to the local field potential (Denker et al., 2011; Kellyet al., 2010; Silva et al., 1991). But oscillations are more than simplythe sum of presynaptic spikes: Many intra- and inter-neuronal signalscontribute to oscillations (Buzsáki et al., 2012), including dynamicintra-laminar and cortical-thalamic interactions (Hindriks and vanPutten, 2013; Jones et al., 2009). Due to intrinsic membrane properties,individual neurons can be “tuned” to specific frequency bands, suchthat theirfiring probability depends both on thephase and the frequencyof the local field potential, a phenomenon known as spike-field coher-ence (Fries et al., 2002; Jacobs et al., 2007; Lepage et al., 2013). Onemajor limitation common to these studies is that they typically donot re-cord EEG as it is used in humans (5–10 mm diameter ring electrodesplaced outside the head); thus it is not clear to what extent the


ED P

RO

OF

relationship between neurophysiological events such as spiking is relat-ed to the EEG signal recorded in humans.

A better understanding of the neurophysiological processes that un-derlie the time–frequency features observed in scalp-recorded M/EEGwould require a two-step approach. The first step would involve simul-taneousmicroscopic andmacroscopic recordings to characterize the ac-tivity at the level of individual neurons and populations of neuronsduring different scalp-recorded EEG events that have been linked to cog-nitive processes. Such research should be done in different brain regionsusing a variety of cognitive tasks, because the mechanisms underlying,for example, visual cortex alpha and prefrontal cortex alpha may be dif-ferent. Most importantly, for thefindings to be relevant to human cogni-tive electrophysiology, the research would have to use electrodes thatare used in humans. The second stepwould involve causal interventionsto test specific hypotheses about the relationship between neurophysi-ological phenomena and scalp-recorded EEG signals. Optogenetics maybe a promising approach (Fenno et al., 2011; LaLumiere, 2011), becausespecific types of neurons can be activated using time courses specifiedby the researcher. For example, it was shown that activating fast-spiking inhibitory interneurons can increase gamma-band oscillationsin local field potentials, which in turn facilitate sensory perception(Cardin et al., 2009) and suppress lower frequency oscillation power(Sohal et al., 2009). Whether and how these local changes would bemeasurable at the level of scalp EEG remain to be determined.

Challenge 3: Reconciling the many diverse spatial filters used incognitive electrophysiology

Spatialfilters useweighted combinations of electrode (or sensor) ac-tivity to isolate features of thedata thatmay bedifficult to observe in thespatially unfiltered data. The spatial filters that are commonly used incognitive electrophysiology include independent components analysis,principle components analysis, the surface Laplacian (also sometimescalled current source density, current scalp density, or dural imaging),single dipole fitting, static distributed source imaging such as LORETA,and adaptive distributed source imaging such as beamforming. Ofcourse, it is also very common not to apply any spatial filters, and in-stead to analyze the spatially unfiltered electrode- or sensor-level data.

Spatial filters were not widely used in the history of cognitive elec-trophysiology. However, as the algorithmic development of spatial fil-ters improves, and as M/EEG recording technology (e.g., more than100 electrodes) allows higher quality recordings with more electrodes,spatial filters will become increasingly important and commonly used.There are three main advantages of using spatial filters (Cohen, 2014):they allow improved localization of activity to topographical or brain re-gions, they facilitate the appropriateness and interpretation of connec-tivity analyses, and they can reveal features of the data that are notreadily visible in the spatially unfiltered data (analogously, gamma-band oscillations can be difficult to observe in EEG data without tempo-ral filtering).

The issues we raise here are that different spatial filters can be qual-itatively different from each other, highlight different features of thedata, and have very different sets of parameters and assumptions. Anexample can be seen in Fig. 1, which shows data from one subject(taken from Cohen and Ridderinkhof, 2013) that were analyzed usingthree different spatial filters (surface Laplacian, independent compo-nents analysis, and beamforming) and no spatial filter. Both commonal-ities and divergences can be seen in the results.

There is no clear consensus regarding which spatial filters shouldbe used in which situations, which analysis and brain-functional as-sumptions are made by different kinds of spatial filters, which param-eter settings are appropriate for which kinds of data, and, importantly,which kinds of interpretations are valid and appropriate for each spa-tial filter. The qualitative differences among spatial filters, and the lackof clear understanding and consensus concerning the various spatialfilters, hinder comparisons and meta-analyses, and cause confusion




UNCO

RRECTED P

RO

OF

A) Spatially unfiltered

0 200 400 600 8002

4

8

15

30

60

0 200 400 600 8002

4

8

15

30

60

0 200 400

dB change from

baseline

600 800

-2

+2

2

4

8

15

30

60

Time (ms)

Fre

qu

ency

(H

z)

Time (ms)

Fre

qu

ency

(H

z)

Time (ms)

Fre

qu

ency

(H

z)

Time (ms)

Fre

qu

ency

(H

z)

0 200 400 600 800

4

8

15

30

60

B) Scalp Laplacian

C) Independent components analysis

D) Source space (via LCMV beamforming)

Fig. 1. Illustration of results of different spatial filters applied to the same single-subject dataset. The right-hand panel shows a time–frequency power plot from electrode FC4 for panels Aand B (see pink electrode), an independent component that was selected based on its topographical distribution (panel C), and a single voxel from a distributed source imaging analysis(LCMV beamforming; see green cross-hairs in panel D). Although there are clearly some similarities, for example, in the ~200–500 ms increase in theta-band power, there are also differ-ences in topography and time–frequency features, such as the ~15–50 Hz relative power suppressions.


Please cite this article as: Cohen, M.X., Gulbinaite, R., Five methodological challenges in cognitive electrophysiology, NeuroImage (2013), http://dx.doi.org/10.1016/j.neuroimage.2013.08.010



T

311

312

313

314

315

316

317

318

319

320

321

322

323

324

325

326

327

328

329

330

331

332

333

334

335

336

337

338

339

340

341

342

343

344

345

346

347

348

349

350

351

352

353

354

355

356

357

358

359

360

361

362

363

364

365

366

367

368

369

370

371

372

373

374

375

376

377

378

379

380

381

382

383

384

385

386

387

388

389

390

391

392

393

394

395

396

397

398

399

400

401

402

403

404

405

406

407

408

409

410

411

412

413

414

415

416

417

418

419

420

421

422

423

424

425

426

427

428

429

430Q3

431

432

433

434

435

436

437

438

439


UNCO

RREC

or misinterpretations among scientists who are not experts in using orinterpreting results from spatially filtered data. Indeed, many discus-sions about this issue seem to occur more often in informal discussionsthan in the scientific literature, making it difficult for non-experts toknow which spatial filters to use or how to evaluate results of differentspatial filters.

The challenge, therefore, is to build a consensus for the situations inwhich each spatial filter is appropriate (including, for example, the kindof data and expected distribution of neural generators), and the kinds ofinterpretations that are appropriate to make from different spatial fil-ters. Reviewing the assumptions and practical implementations of allspatial filters is a task too great and too lengthy for this review. Here,as an example of the differences in spatial filters, we highlight two spe-cific filters that are used in connectivity analyses: Beamforming and thesurface Laplacian. Beamforming is an adaptive spatial filter that utilizesthe covariance between a forward model (an estimate of the scalp-measured signal given activity in a location in the brain) and the ob-served electrode-level data to estimate the potential brain sources ofscalp-observed signals. Beamforming relies on several assumptionsabout brain and head shape, the strength of correlated sources, theorientations of dipoles, and, for EEG, the conductances of differenthead tissues. The surface Laplacian is a spatial band-pass filter thatattenuates spatial low-frequency components, which often reflectvolume-conduced sources. The surface Laplacian relies on few assump-tions and parameters (except spatial smoothness), but is limited toscalp-level inferences. Despite their differences, both spatial filtershave been recommended for inter-regional connectivity analyses(Schoffelen and Gross, 2009; Srinivasan et al., 2007), but, to our knowl-edge, these have not been directly compared. It is unclear from the liter-ature which of these two spatial filters is best for connectivity analysesin which circumstances.

This challenge can be addressed in two ways. First, there can bededicated methods papers that compare results of different spatial fil-ters using simulated data as well as real data that likely have distribut-ed brain generators (an online dataset has recently been madeavailable specifically for this purpose; Aine et al., 2012). Although sim-ulated data are useful because the ground truth is known, they oftenlack realistic characteristics of EEG data, including noise, travelingwaves, and brain region interactions, and may be biased towards spe-cific methods. Real datasets, on the other hand, are likely to providemore insights into practicalities of spatial filters, but it may be difficultto knowwhat the “true” result is when different spatial filters producedifferent results. Much of the extant literature comparing spatial fil-ters focuses on comparing parameter settings for the same filter, orcomparing different algorithms for the same type of filter (Dalal et al.,2008; Hansen et al., 2010; Hauk, 2004; Michel et al., 2004; Nunezet al., 1993; Quraan and Cheyne, 2010; Sekihara and Nagarajan, 2008).These investigations are extremely useful forwithin-filter comparisons;future studies should additionally focus on comparisons across spatialfilters.

The second way to address this challenge is in studies that arecontent-oriented (as opposed to being specifically methods-oriented).Here, analyses can be performed with a few different spatial filters todetermine whether and how the results depended on the specific spa-tial filter. It is likely that qualitative differences among spatial filterswill arise in more detailed analyses, such as inter-regional frequencyband-specific connectivity, or when considering complex frequencypatterns (Fig. 1). Two examples from the literature highlight that in dif-ferent situations, spatial filters can have little (Cohen, 2011) or signifi-cant (Rivet et al., 2012) effects on the results.

Of course, the spatial accuracy of EEG and MEG is limited comparedto fMRI or invasive recordings. Nonetheless, there is considerably morespatial information available in high-density recordings (N100 elec-trodes) than in the low-density recordings (b33 electrodes). As the spa-tial information available in the M/EEG data continues to increase, andas an understanding of distributed network functioning becomes


increasingly relevant in cognitive electrophysiology, building a consen-sus about spatial filters becomes an important challenge to be met.

ED P

RO

OF

Challenge 4: Characterizing multi-scale interactions inneuroelectric activity

Spatial and temporal multiscale interactions are thought to be a de-fining feature of the brain (Fig. 2), and a critical principle that underliescognitive functions and consciousness (Breakspear and Stam, 2005; LeVan Quyen, 2011; Varela et al., 2001). The cognitive electrophysiologyliterature spans a broad range of spatiotemporal scales: Spatial scalesrange from single unit firings (a very small spatial scale) to local fieldpotentials (a relatively small spatial scale) to source-reconstructedEEG or MEG activity (a relatively medium spatial scale) to low-spatial-resolution EEG (i.e., fewer than 33 electrodes; a relatively large spatialscale); and temporal scales range frommilliseconds (e.g., spike timing)to minutes (e.g., resting-state). In some cases, spatial and temporalscales may be related. For example, in the motor system, beta-band ac-tivity appears to be relatively spatially widespread while gamma-bandactivity appears to be more spatially restricted (Miller et al., 2007).However, time and space are not necessary linked, and the idea that,for example, lower temporal frequencies correspond to larger brain net-works (von Stein and Sarnthein, 2000) is a useful conceptualization butis not fully consistent with empirical findings. For example, large-scalebrain networks can also be entrained in the gamma band (Doesburget al., 2008; Pesaran et al., 2008; Siegel et al., 2008).

Despite the theoretical importance of multiscale interactions forbrain function, most individual cognitive electrophysiology studies uti-lize only one spatial and temporal scale. This means that comparisonsacross spatial and temporal scales are often made loosely and qualita-tively, across different studies. The challenge, therefore, is to determineto what extent dynamics at different spatial and temporal scales are re-lated to each other, and to determine how these multiscale dynamicsare in turn related to cognitive processes.

A better understanding of multiscale interactions is important fortwo reasons. First, the mechanisms that govern brain function at onespatiotemporal scale and in one brain region may not generalize toother spatiotemporal scales and other brain regions. Second, mecha-nisms that are observed at one spatiotemporal scale may be difficultor impossible to explain without understanding mechanisms at largeror smaller spatiotemporal scales. An example of this latter phenomenonis that the timing of an individual action potential may seem randomuntil discovering that that action potential is locked to a specific phaseof the local field potential.

There are several possible approaches to address this challenge.Most importantly, addressing this challenge will require a widening ofperspectives on the role of multiscale interactions in brain function.Most scientists would agree that multiscale interactions are importantfor brain function, and yet few incorporate these ideas into empirical re-search. For example, in EEG research it is typical to filter out activity atboth low (b1 Hz) and high (N40 Hz) frequencies; thus, potentially rel-evant multiscale temporal interactions are not considered (Palva andPalva, 2012).

Second, theories will need to be developed in order to make predic-tions, guide analyses, and interpret results concerning multiscale inter-actions (see also Challenge 5 concerning the role of theories in cognitiveelectrophysiology). An illustration of a theory that makes testable pre-dictions for multiscale temporal interactions comes from workingmemory studies (Axmacher et al., 2010; Jensen and Lisman, 1998;Sauseng et al., 2009), in which the number of gamma cycles observedin a theta cycle (e.g., cross-frequency coupling; a temporal multiscalephenomenon) is thought to vary as a function of memory load or mem-ory capacity. What specific predictions about multiscale interactionscould bemade regarding other cognitive processes such as action mon-itoring, learning, decision-making, language, and social imitation?




RRECTED P

RO

OF

440

441

442

443

444

445

446

447

448

449

450

451

452

453

454

455

456

457

458

459

460

461

462

463

464

465

466

467

468

469

470

471

472

473

474

I

II/III

IV

V/VI

I

II/III

IV

V/VI

x100 neuronsSingle unitA)

B)

C)

D)

E)

F)

G)

Cortical column/ensemble

LFP

Action potentialsBursting

EEG α EEG

β−γ LFP

Connectivity

Cortical patch

Large-scale interactions

x103 neurons

x104 neurons

x106 neurons

electrode

Grandma??

20 ms 20 ms

100 ms 100 ms

300 ms 300 ms

1000 ms

Fig. 2. To understand the information content of neural dynamics at one spatiotemporal scale, it might be necessary to consider the neural dynamics at other spatiotemporal scales. (A) Asingle neuron emits action potentials in response to a stimulus. The information contained in action potential timing may depend on the simultaneous activity of the neural ensemble inwhich that neuron is embedded (B). The collective activity of the ensemble can be measured as the local field potential (LFP), which can encode information in different frequenciessimultaneously, and can also functionally connect neighboring neural ensembles. (C) These synchronous neural ensembles form cortical “patches” that produce electrical fields largeenough to be recorded with the EEG. (D) Large-scale brain networks become synchronized, for example, during top-down control processes such as attention. The processes depictedin panels A–D are often termed “bottom-up,” because the spatiotemporal scales are increasing. These same pathways are also used for “top-down” processes, as illustrated in panelsD–G. Top-down processes such as attention can modulate neural activity at cortical patches, (F) which in turn can regulate activity in neural ensembles, (G) which in turn can modulateactivity of single neurons. Sub-neuron-level dynamics (e.g., synapses) and supra-brain dynamics (e.g., inputs from other body systems such as digestion, breathing, and heart rate) are notdepicted here, but are also relevant for brain function and neural computation.


UNCO

Third, this challenge will require developments in data acquisitiontechniques. Most brain imaging technologies are designed to measureone spatial scale; studying spatialmultiscale relationships often requiresspecialized equipment thatmay be difficult to access or thatmay requireconsiderable technical expertise (for example, simultaneous single-cellelectrophysiology and fMRI). Multiscale interactions can also be mea-sured by combining methodological approaches at different spatialscales, such as combined EEG–fMRI (Debener et al., 2005; Jann et al.,2012; Scheeringa et al., 2011). In this case, the fMRI signal has higherspatial resolution while EEG has higher temporal precision.

Relatedly, there will need to be developments in the data analysesthat are used to quantify and conceptualize multiscale interactions. Atpresent, multiscale temporal interactions are most often assessedthrough cross-frequency coupling (Lisman and Jensen, 2013; Youngand Eggermont, 2009). Cross-frequency coupling has been related to in-formation processing schemes (Lisman, 2005), and there are many em-pirical demonstrations of cross-frequency coupling characteristics beingmodulated by cognitive task demands (Canolty and Knight, 2010;Cohen and van Gaal, 2012; Kayser et al., 2012; Voytek et al., 2010;


Young and Eggermont, 2009). Graph theory provides some useful met-rics for characterizing how local activity can be modulated by network-level dynamics (e.g., Eldawlatly et al., 2009). In some cases, both tempo-ral and spatial multiscale interactions can be examined simultaneously(Le Van Quyen et al., 2013; van der Meij et al., 2012).

Challenge 5: Developing neurophysiologically groundedpsychological theories

Theories facilitate scientific development. They provide frameworksfor generating new experiments and hypotheses, interpreting results,and comparing results acrossmethodologies, species, and levels of anal-ysis (Abbott, 2008). Theories that provide an interface between psycho-logical constructs and neural dynamics are particularly important,because the gap between, for example, the action potentials of a neuronin visual cortex and the subjective report of having seen a briefly flashedvisual stimulus, is quite large; good theories can help bridge these kindsof distances.




T

475

476

477

478

479

480

481

482Q4

483

484

485

486

487

488

489

490

491

492

493

494

495

496

497

498

499

500

501

502

503

504

505

506

507

508

509

510

511

512

513

514

515

516

517

518

519

520

521

522

523

524

525

526

527

528

529

530

531

532

533

534

535

536

537

538

539

540

541

542

543

544

545

546

547

548

549

550

551

552

553

554

555

556

557

558

559

560

561

562

563

564

565

566

567

568

569

570

571

572

573

574

575576577578579580581582583584585586Q6587588589590591592593594595596597598599600601602603604605606607

i Thanks to Ole Jensen for this reference.


UNCO

RREC

The challenge is to develop and expand theories that can account forand make specific predictions regarding cognitive processes and theneurophysiological dynamics that accompany those processes (includ-ing frequency band-specific responses, connectivity, etc.). At present,many psychological theories that make predictions for brain activity as-sume that cognitive functions are localized to specific brain regions, andthat the operation of a cognitive function can be measured as “activa-tion” of that brain region (see Challenge 1 for a discussion of the termactivation). These types of “cognitive-level” models have been impor-tant for the development of cognitive science and cognitive neurosci-ence, and we do not intend to criticize this approach generally.However, it has become clear that at the electrophysiological level, pre-dictions of cognitive models that assume a unidimensional “activation/deactivation” scale for brain activity become neither confirmable norfalsifiable.

The under-specification of neurophysiological dynamics in psycho-logical theories presents both theoretical and statistical concerns. Theo-retically, predictions concerning neurophysiological implementationsbecome difficult to confirm or disprove, in part because the searchspace in which a feature of data can be post-hoc labeled “activation” islarge. For example, imagine a theory that predicts increased activationin a brain region, and two experiments are carried out to test this predic-tion. One experiment finds an increase in theta-band power, decrease inalpha-band power, and increased inter-regional connectivity; the otherexperiment finds no change in theta- or alpha-band power, increases inbeta- and gamma-band power, and increases in theta–gamma coupling.The results of these two experiments are quite different; which experi-ment is more consistent with the theory? In this case, the answer isneither, because the theory is too under-specified to help resolve theinter-experiment inconsistencies.

This theoretical concern is further compounded by a statistical con-cern: The high-dimensional space of electrophysiology time–frequencyresults provides a large number of statistical evaluations, which requireappropriately strict statistical correction for multiple comparisons.Without better theories that make more precise predictions, statisticalsensitivity may be too low to detect subtle but true findings whencorrecting for multiple comparisons over time, frequency, and space(Ashton, 2013). Furthermore, without theoretical guidance, the re-searcher might fail to recognize a true finding when it is present in thedata.

Addressing this challenge will require a deeper integration betweencognitive psychology, electrophysiology, and biophysical principles.Fortunately, making this connection becomes easier, with develop-ments in neuroscience, computing, and electrophysiology. For example,there are now simplified computational models that contain sufficientbiophysical plausibility to simulate local field potential and EEG data,such as neural mass and neural field models (Babajani-Feremi andSoltanian-Zadeh, 2010; Deco et al., 2008; Nguyen Trong et al., 2012).Furthermore, specialized software for simulating small- and large-scale networks of biophysically plausible neurons are under constantdevelopment (e.g., “NEST”: Gewaltig and Diesmann, 2007; “Brian”:Goodman and Brette, 2008; “NEURON”: Hines and Carnevale, 1997). Al-thoughmanybiophysicalmodels do not exhibit cognitive behaviors andmany “cognitive-level” theories do not contain biophysical plausibility,there are no longer major scientific or practical stymies that preventdeeper links between cognitive and electrophysiology. The develop-ment of theories that account for both neurophysiological and psycho-logical phenomena may prove to be transformative for the field ofcognitive electrophysiology.

There are several theories that bridge cognitive processes and physio-logical mechanisms, and therefore meet this challenge. Following arethree examples. First, one recent theory (Jensen et al., 2012) incorporatesboth the putative biophysical mechanisms of alpha oscillations and thecognitive consequences in terms of when salient or attended visualstimuli can be processed. Second, there are theories ofworkingmemorythat explain how reverberatory activity in biophysically plausible neural


ED P

RO

OF

networks can be used to maintain goal-relevant representations thatare robust against distracters (Amit and Brunel, 1997; Compte et al.,2000). Third, theories of simple perceptual decision-making provideputative biophysical circuit mechanisms underlying simple sensorydiscrimination (Wang, 2012). These are not the only examples; othermodels contain some biophysical plausibility and can also explainsome aspects of sensory, perceptual, and cognitive phenomena (Ardidet al., 2010; Buia and Tiesinga, 2006; Corchs and Deco, 2002; Wieckiand Frank, 2013). Nonetheless, many cognitive processes and theirneurophysiological correlates are not accounted for by current theories.Indeed, the claim that “electroencephalography is still mainly an empir-ical science” (Zhadin, 1984) is as true now as it was 30 years ago.

New horizons in cognitive electrophysiology

This is an exciting time to be a cognitive electrophysiologist. Devel-opments in M/EEG recording equipment provide increasingly higherquality data, and developments in computing technology allow increas-ingly sophisticated analyses of those data. Furthermore, developmentsin neuroscience, neurobiology, and biophysical modeling allow re-searchers to link their scalp-recorded findings (in particular, findingssuggesting large-scale oscillatory dynamics) to neurophysiologicalmechanisms with a level of detail not previously possible in the historyof non-invasive EEG. The next few decades are likely to see significantand possibly qualitative improvements in understanding how cognitivefunctions are produced by multiscale integrative electrophysiologicaldynamics. These developments will likely outpace the developmentsover the previous decades, considering for example that the link be-tween inter-regional spectral coherence and learning was alreadyknown in the 1970's (Busk and Galbraith, 1975)i.

To be sure, there is much yet to learn. Perhaps in the year 2033 we'lllook back andmarvel at howprimitive and uninformed our theories andanalyses were in the year 2013. Our aim in this paper was to point outsome of the methodological challenges that, if met, will help cognitiveelectrophysiology develop as a science into the 21st century.

References

Abbott, L.F., 2008. Theoretical neuroscience rising. Neuron 60, 489–495.Aine, C.J., Sanfratello, L., Ranken, D., Best, E., MacArthur, J.A., Wallace, T., Gilliam, K.,

Donahue, C.H., Montaño, R., Bryant, J.E., Scott, A., Stephen, J.M., 2012. MEG-SIM: aweb portal for testing MEG analysis methods using realistic simulated and empiricaldata. Neuroinformatics 10, 141–158.

Amit, D.J., Brunel, N., 1997. Dynamics of a recurrent network of spiking neurons beforeand following learning. Netw. Comput. Neural Syst. 8, 373–404.

Ardid, S., Wang, X.-J., Gomez-Cabrero, D., Compte, A., 2010. Reconciling coherent oscilla-tion with modulation of irregular spiking activity in selective attention: gamma-range synchronization between sensory and executive cortical areas. J. Neurosci.Off. J. Soc. Neurosci. 30, 2856–2870.

Ashton, J.C., 2013. Experimental power comes from powerful theories— the real problemin null hypothesis testing. Nat. Rev. Neurosci. 14, 585-585.

Axmacher, N., Henseler, M.M., Jensen, O., Weinreich, I., Elger, C.E., Fell, J., 2010. Cross-frequency coupling supports multi-item working memory in the human hippocam-pus. Proc. Natl. Acad. Sci. 107, 3228–3233.

Babajani-Feremi, A., Soltanian-Zadeh, H., 2010. Multi-area neural mass modeling of EEGand MEG signals. NeuroImage 52, 793–811.

Breakspear, M., Stam, C.J., 2005. Dynamics of a neural system with a multiscale architec-ture. Philos. Trans. R. Soc. B-Biol. Sci. 360, 1051–1074.

Bruns, A., Eckhorn, R., Jokeit, H., Ebner, A., 2000. Amplitude envelope correlation detectscoupling among incoherent brain signals. Neuroreport 11, 1509–1514.

Buia, C., Tiesinga, P., 2006. Attentional modulation of firing rate and synchrony in a modelcortical network. J. Comput. Neurosci. 20, 247–264.

Burgess, A.P., 2012. Towards a unified understanding of event-related changes in the EEG:the fireflymodel of synchronization through cross-frequency phasemodulation. PLoSOne 7, e45630.

Burke, J.F., Long, N.M., Zaghloul, K.A., Sharan, A.D., Sperling, M.R., Kahana, M.J., 2013.Human Intracranial High-frequency Activity Maps Episodic Memory Formation inSpace and Time. Neuroimage.

Bush, P., Sejnowski, T., 1996. Inhibition synchronizes sparsely connected cortical neuronswithin and between columns in realistic network models. J. Comput. Neurosci. 3,91–110.




T

608609610611612613614615616617618619620621622623624625626627628629630631632Q7633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677Q8678679680681682683684685686687688689690691692693

694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779


UNCO

RREC

Busk, J., Galbraith, G.C., 1975. EEG correlates of visual-motor practice in man.Electroencephalogr. Clin. Neurophysiol. 38, 415–422.

Buzsáki, G., Wang, X.-J., 2012. Mechanisms of gamma oscillations. Annu. Rev. Neurosci.35, 203–225.

Buzsáki, G., Anastassiou, C.A., Koch, C., 2012. The origin of extracellular fields and currents—EEG, ECoG, LFP and spikes. Nat. Rev. Neurosci. 13, 407–420.

Canolty, R.T., Knight, R.T., 2010. The functional role of cross-frequency coupling. TrendsCogn. Sci. 14, 506–515.

Cardin, J.A., Carlén, M., Meletis, K., Knoblich, U., Zhang, F., Deisseroth, K., Tsai, L.-H., Moore,C.I., 2009. Driving fast-spiking cells induces gamma rhythm and controls sensory re-sponses. Nature 459, 663–667.

Cavanagh, J.F., Cohen, M.X., Allen, J.J.B., 2009. Prelude to and resolution of an error: EEGphase synchrony reveals cognitive control dynamics during action monitoring.J. Neurosci. Off. J. Soc. Neurosci. 29, 98–105.

Coenen, A.M., 1995. Neuronal activities underlying the electroencephalogram and evokedpotentials of sleeping and waking: implications for information processing. Neurosci.Biobehav. Rev. 19, 447–463.

Cohen, M.X., 2011. Error-related medial frontal theta activity predicts cingulate-relatedstructural connectivity. NeuroImage 55, 1373–1383.

Cohen, M.X., 2014. Analyzing Neural Time Series Data: Theory and Practice. MIT Press.Cohen, M.X., Ridderinkhof, K.R., 2013. EEG source reconstruction reveals frontal–parietal

dynamics of spatial conflict processing. PLoS One 8, e57293.Cohen, M.X., van Gaal, S., 2012. Dynamic interactions between large-scale brain networks

predict behavioral adaptation after perceptual errors. Cereb. Cortex (New York N1991).

Compte, A., Brunel, N., Goldman-Rakic, P.S., Wang, X.J., 2000. Synaptic mechanisms andnetwork dynamics underlying spatial working memory in a cortical networkmodel. Cereb. Cortex 10, 910–923 (New York N 1991).

Corchs, S., Deco, G., 2002. Large-scale neural model for visual attention: integration of ex-perimental single-cell and fMRI data. Cereb. Cortex 12, 339–348 (New York N 1991).

Crone, N.E., Korzeniewska, A., Franaszczuk, P.J., 2011. Cortical γ responses: searching highand low. Int. J. Psychophysiol. 79, 9–15.

Dalal, S.S., Guggisberg, A.G., Edwards, E., Sekihara, K., Findlay, A.M., Canolty, R.T., Berger,M.S., Knight, R.T., Barbaro, N.M., Kirsch, H.E., Nagarajan, S.S., 2008. Five-dimensionalneuroimaging: localization of the time–frequency dynamics of cortical activity.NeuroImage 40, 1686–1700.

David, O., Kilner, J.M., Friston, K.J., 2006. Mechanisms of evoked and induced responses inMEG/EEG. NeuroImage 31, 1580–1591.

Debener, S., Ullsperger, M., Siegel, M., Fiehler, K., von Cramon, D.Y., Engel, A.K., 2005. Trial-by-trial coupling of concurrent electroencephalogram and functional magnetic reso-nance imaging identifies the dynamics of performance monitoring. J. Neurosci. Off.J. Soc. Neurosci. 25, 11730–11737.

Deco, G., Jirsa, V.K., Robinson, P.A., Breakspear, M., Friston, K., 2008. The dynamic brain:from spiking neurons to neural masses and cortical fields. PLoS Comput. Biol. 4,e1000092.

Denker,M., Roux, S., Lindén, H., Diesmann,M., Riehle, A., Grün, S., 2011. The local field poten-tial reflects surplus spike synchrony. Cereb. Cortex 21, 2681–2695 (New York N 1991).

Doesburg, S.M., Roggeveen, A.B., Kitajo, K., Ward, L.M., 2008. Large-scale gamma-bandphase synchronization and selective attention. Cereb. Cortex 18, 386–396 (NewYork N 1991).

Donner, T.H., Siegel, M., 2011. A framework for local cortical oscillation patterns. TrendsCogn. Sci. 15, 191–199.

Eldawlatly, S., Jin, R., Oweiss, K.G., 2009. Identifying functional connectivity in large-scaleneural ensemble recordings: a multiscale data mining approach. Neural Comput. 21,450–477.

Engel, A.K., König, P., Kreiter, A.K., Schillen, T.B., Singer, W., 1992. Temporal coding in thevisual cortex: new vistas on integration in the nervous system. Trends Neurosci. 15,218–226.

Fenno, L., Yizhar, O., Deisseroth, K., 2011. The development and application of optogenetics.Annu. Rev. Neurosci. 34, 389–412.

Fries, P., Schröder, J.-H., Roelfsema, P.R., Singer, W., Engel, A.K., 2002. Oscillatory neuronalsynchronization in primary visual cortex as a correlate of stimulus selection.J. Neurosci. Off. J. Soc. Neurosci. 22, 3739–3754.

Gardiner, Java, 1993. Recognising and remembering. In: Collins, A., Gathercole, S., Conway,M., Morris, P. (Eds.), Theories of Memory. Erlbaum, Hillsdale, NJ, pp. 163–188.

Gewaltig, M.-O., Diesmann, M., 2007. NEST (NEural Simulation Tool). Scholarpedia 2,1430.

Goodman, D., Brette, R., 2008. Brian: a simulator for spiking neural networks in Python.Front. Neuroinform. 2, 5.

Hansen, P., Kringelbach, M., Salmelin, R., 2010. MEG: An Introduction to Methods: An In-troduction to Methods. Oxford University Press.

Hauk, O., 2004. Keep it simple: a case for using classical minimum norm estimation in theanalysis of EEG and MEG data. NeuroImage 21, 1612–1621.

He, B.J., Zempel, J.M., Snyder, A.Z., Raichle, M.E., 2010. The temporal structures and func-tional significance of scale-free brain activity. Neuron 66, 353–369.

Hindriks, R., van Putten, M.J.A.M., 2013. Thalamo-cortical mechanisms underlying changesin amplitude and frequency of human alpha oscillations. NeuroImage 70, 150–163.

Hines, M.L., Carnevale, N.T., 1997. The NEURON simulation environment. Neural Comput.9, 1179–1209.

Jacobs, J., Kahana, M.J., Ekstrom, A.D., Fried, I., 2007. Brain oscillations control timingof single-neuron activity in humans. J. Neurosci. Off. J. Soc. Neurosci. 27,3839–3844.

Jann, K., Federspiel, A., Giezendanner, S., Andreotti, J., Kottlow, M., Dierks, T., Koenig, T.,2012. Linking brain connectivity across different time scales with electroencephalo-gram, functional magnetic resonance imaging, and diffusion tensor imaging. BrainConnect 2, 11–20.


ED P

RO

OF

Jensen, O., Lisman, J.E., 1998. An oscillatory short-termmemory buffer model can accountfor data on the Sternberg task. J. Neurosci. 18, 10688–10699.

Jensen, O., Bonnefond, M., VanRullen, R., 2012. An oscillatory mechanism for prioritizingsalient unattended stimuli. Trends Cogn. Sci. 16, 200–206.

Jones, S.R., Pritchett, D.L., Sikora, M.A., Stufflebeam, S.M., Hämäläinen, M., Moore, C.I.,2009. Quantitative analysis and biophysically realistic neural modeling of theMEGmu rhythm: rhythmogenesis and modulation of sensory-evoked responses.J. Neurophysiol. 102, 3554–3572.

Kayser, C., Ince, R.A.A., Panzeri, S., 2012. Analysis of slow (theta) oscillations as a potentialtemporal reference frame for information coding in sensory cortices. PLoS Comput.Biol. 8, e1002717.

Kelly, R.C., Smith, M.A., Kass, R.E., Lee, T.S., 2010. Local field potentials indicate networkstate and account for neuronal response variability. J. Comput. Neurosci. 29, 567–579.

Kraut, M.A., Arezzo, J.C., Vaughan Jr., H.G., 1985. Intracortical generators of the flash VEP inmonkeys. Electroencephalogr. Clin. Neurophysiol. 62, 300–312.

Lachaux, J.P., Rodriguez, E., Martinerie, J., Varela, F.J., 1999. Measuring phase synchrony inbrain signals. Hum. Brain Mapp. 8, 194–208.

LaLumiere, R.T., 2011. A new technique for controlling the brain: optogenetics and itspotential for use in research and the clinic. Brain Stimul. 4, 1–6.

Le Van Quyen, M., 2011. The brainweb of cross-scale interactions. New Ideas Psychol. 29,57–63.

Le Van Quyen, M., Botella-Soler, V., Valderrama, M., 2013. Neuronal oscillations scale upand scale down the brain dynamics. In: Meyer Pesenon, Misha (Ed.), Multiscale Anal-ysis and Nonlinear Dynamics, Reviews of Nonlinear Dynamics and Complexity.Wiley-VCH, Weinheim Germany, pp. 205–214.

Lepage, K.Q., Gregoriou, G.G., Kramer, M.A., Aoi, M., Gotts, S.J., Eden, U.T., Desimone, R.,2013. A procedure for testing across-condition rhythmic spike-field associationchange. J. Neurosci. Methods 213, 43–62.

Lisman, J., 2005. The theta/gamma discrete phase code occuring during the hippocampalphase precession may be a more general brain coding scheme. Hippocampus 15,913–922.

Lisman, J.E., Jensen, O., 2013. The θ–γ neural code. Neuron 77, 1002–1016.Lopes da Silva, F., van Rotterdam, A., 1982. Biophysical aspects of EEG andMEG generation.

In: Niedermeyer, E., Lopes da Silva, F. (Eds.), Electroencephalography: Basic Principles,Clinical Applications, and Related Fields. LippincottWilliams &Wilkins, BaltimoreMD.

Makeig, S., 1993. Auditory event-related dynamics of the EEG spectrum and effects ofexposure to tones. Electroencephalogr. Clin. Neurophysiol. 86, 283–293.

Manning, J.R., Jacobs, J., Fried, I., Kahana, M.J., 2009. Broadband shifts in local field poten-tial power spectra are correlated with single-neuron spiking in humans. J. Neurosci.Off. J. Soc. Neurosci. 29, 13613–13620.

McLelland, D., Paulsen, O., 2009. Neuronal oscillations and the rate-to-phase transform:mechanism, model and mutual information. J. Physiol. 587, 769–785.

Michel, C.M., Murray, M.M., Lantz, G., Gonzalez, S., Spinelli, L., Grave de Peralta, R., 2004.EEG source imaging. Clin. Neurophysiol. Off. J. Int. Fed. Clin. Neurophysiol. 115,2195–2222.

Miller, K.J., Leuthardt, E.C., Schalk, G., Rao, R.P.N., Anderson, N.R., Moran, D.W., Miller, J.W.,Ojemann, J.G., 2007. Spectral changes in cortical surface potentials during motormovement. J. Neurosci. Off. J. Soc. Neurosci. 27, 2424–2432.

Miller, K.J., Zanos, S., Fetz, E.E., den Nijs, M., Ojemann, J.G., 2009. Decoupling the corticalpower spectrum reveals real-time representation of individual finger movements inhumans. J. Neurosci. Off. J. Soc. Neurosci. 29, 3132–3137.

Nguyen Trong, M., Bojak, I., Knösche, T.R., 2012. Associating spontaneous with evokedactivity in a neural mass model of visual cortex. NeuroImage 66C, 80–87.

Nunez, P.L., Silberstein, R.B., Cadusch, P.J., Wijesinghe, R., 1993. Comparison of high reso-lution EEG methods having different theoretical bases. Brain Topogr. 5, 361–364.

Palva, J.M., Palva, S., 2012. Infra-slow fluctuations in electrophysiological recordings,blood-oxygenation-level-dependent signals, and psychophysical time series.NeuroImage 62, 2201–2211.

Pesaran, B., Nelson, M.J., Andersen, R.A., 2008. Free choice activates a decision circuitbetween frontal and parietal cortex. Nature 453, 406–409.

Pfurtscheller, G., 1992. Event-related synchronization (ERS): an electrophysiologicalcorrelate of cortical areas at rest. Electroencephalogr. Clin. Neurophysiol. 83, 62–69.

Pfurtscheller, G., Lopes da Silva, F.H., 1999. Event-related EEG/MEG synchronization anddesynchronization: basic principles. Clin. Neurophysiol. Off. J. Int. Fed. Clin. Neurophysiol.110, 1842–1857.

Quraan, M.A., Cheyne, D., 2010. Reconstruction of correlated brain activity with adaptivespatial filters in MEG. NeuroImage 49, 2387–2400.

Ray, S., Crone, N.E., Niebur, E., Franaszczuk, P.J., Hsiao, S.S., 2008. Neural correlates of high-gamma oscillations (60–200 Hz) in macaque local field potentials and their potentialimplications in electrocorticography. J. Neurosci. Off. J. Soc.Neurosci. 28, 11526–11536.

Rieke, F., 1999. Spikes: Exploring the Neural Code. University Press Group Limited.Rivet, B., Cecotti, H., Maby, E., Mattout, J., 2012. Impact of spatial filters during sensor

selection in a visual P300 brain–computer interface. Brain Topogr. 25, 55–63.Sauseng, P., Klimesch, W., Heise, K.F., Gruber, W.R., Holz, E., Karim, A.A., Glennon, M.,

Gerloff, C., Birbaumer, N., Hummel, F.C., 2009. Brain oscillatory substrates of visualshort-term memory capacity. Curr. Biol. 19, 1846–1852.

Scheeringa, R., Mazaheri, A., Bojak, I., Norris, D.G., Kleinschmidt, A., 2011. Modulation ofvisually evoked cortical FMRI responses by phase of ongoing occipital alpha oscilla-tions. J. Neurosci. Off. J. Soc. Neurosci. 31, 3813–3820.

Schoffelen, J.-M., Gross, J., 2009. Source connectivity analysis with MEG and EEG. Hum.Brain Mapp. 30, 1857–1865.

Schroeder, C.E., Tenke, C.E., Givre, S.J., Arezzo, J.C., Vaughan Jr., H.G., 1991. Striate corticalcontribution to the surface-recorded pattern-reversal VEP in the alert monkey. VisionRes. 31, 1143–1157.

Sekihara, K., Nagarajan, S.S., 2008. Adaptive Spatial Filters for Electromagnetic Brain Imag-ing. Springer.




780781782783784785786787788789790791792793794795796797798799800801

802803804805806807808809810811812813814815816817818819820821822823

825


Siegel, M., Donner, T.H., Oostenveld, R., Fries, P., Engel, A.K., 2008. Neuronal synchroniza-tion along the dorsal visual pathway reflects the focus of spatial attention. Neuron 60,709–719.

Silva, L.R., Amitai, Y., Connors, B.W., 1991. Intrinsic oscillations of neocortex generated bylayer 5 pyramidal neurons. Science 251, 432–435.

Singh, K.D., 2012. Which “neural activity” do you mean? fMRI, MEG, oscillations and neu-rotransmitters. NeuroImage 62, 1121–1130.

Sohal, V.S., Zhang, F., Yizhar, O., Deisseroth, K., 2009. Parvalbumin neurons and gammarhythms enhance cortical circuit performance. Nature 459, 698–702.

Srinivasan, R., Winter, W.R., Ding, J., Nunez, P.L., 2007. EEG andMEG coherence: measuresof functional connectivity at distinct spatial scales of neocortical dynamics.J. Neurosci. Methods 166, 41–52.

Tallon-Baudry, Bertrand, 1999. Oscillatory gamma activity in humans and its role in objectrepresentation. Trends Cogn. Sci. 3, 151–162.

Tsukada, M., Ichinose, N., Aihara, K., Ito, H., Fujii, H., 1996. Dynamical cell assemblyhypothesis — theoretical possibility of spatio-temporal coding in the cortex. NeuralNetw. 9, 1303–1350.

Tulving, E., 2000. Concepts of memory. The Oxford Handbook of Memory. Oxford Univer-sity Press, pp. 33–43.

Van der Meij, R., Kahana, M., Maris, E., 2012. Phase–amplitude coupling in human electro-corticography is spatially distributed and phase diverse. J. Neurosci. Off. J. Soc.Neurosci. 32, 111–123.

UNCO

RRECT

824


F

Van Steenbergen, H., Band, G.P.H., Hommel, B., 2012. Reward valence modulates conflict-driven attentional adaptation: electrophysiological evidence. Biol. Psychol. 90,234–241.

Varela, F., Lachaux, J.-P., Rodriguez, E., Martinerie, J., 2001. The brainweb: phase synchro-nization and large-scale integration. Nat. Rev. Neurosci. 2, 229–239.

Von Stein, A., Sarnthein, J., 2000. Different frequencies for different scales of corticalintegration: from local gamma to long range alpha/theta synchronization. Int.J. Psychophysiol. 38, 301–313.

Voytek, B., Canolty, R.T., Shestyuk, A., Crone, N.E., Parvizi, J., Knight, R.T., 2010. Shifts ingamma phase–amplitude coupling frequency from theta to alpha over posteriorcortex during visual tasks. Front. Hum. Neurosci. 4, 191.

Wang, X.-J., 2010. Neurophysiological and computational principles of cortical rhythms incognition. Physiol. Rev. 90, 1195–1268.

Wang, X.-J., 2012. Neural dynamics and circuit mechanisms of decision-making. Curr.Opin. Neurobiol. 22, 1039–1046.

Whittingstall, K., Logothetis, N.K., 2009. Frequency-band coupling in surface EEG reflectsspiking activity in monkey visual cortex. Neuron 64, 281–289.

Wiecki, T.V., Frank, M.J., 2013. A computational model of inhibitory control in frontalcortex and basal ganglia. Psychol. Rev. 120, 329–355.

Young, C.K., Eggermont, J.J., 2009. Coupling of mesoscopic brain oscillations: recentadvances in analytical and theoretical perspectives. Prog. Neurobiol. 89, 61–78.

Zhadin,M.N., 1984. Rhythmic processes in the cerebral cortex. J. Theor. Biol. 108, 565–595.

O

ED P

RO




Five methodological challenges in cognitive electrophysiology

Documents

Transcript of Five methodological challenges in cognitive electrophysiology