Frontal-midlinethetafromtheperspectiveofhippocampalthetaDamonJ.Mitchella,NeilMcNaughtona,*,Danny Flanaganb,Ian J.KirkcaDepartmentofPsychologyandCentreforNeuroscience,UniversityofOtago,Dunedin,NewZealandbBrainResearchInstitute,AustinHealth,HeidelbergWest,Melbourne,AustraliacDepartmentofPsychologyandResearchCentreforCognitiveNeuroscience,UniversityofAuckland,NewZealandContents1. Introduction . .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. . 1571.1. Afocusonthetarhythms .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. . 1571.2. Fromrhythmtofunction .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. . 1581.3. Ahippocampalperspective.. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. . 1581.4. Pragmatismversustheory . .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. . 158ProgressinNeurobiology 86(2008)156185ARTI CLE I NFOArticlehistory:Received19 December2007Receivedin revisedform24July2008Accepted3September2008Keywords:ThetaFrontalcortexHippocampusFrontal-midlinethetaRhythmicalslow activityEEGWorkingmemorySpatialnavigationEpisodicmemoryInternalisedattentionMeditationAnxietyAnxiolyticPersonalityABSTRACTElectrical recordings from the surface of the skull have a wide range of rhythmic components. A majortaskofanalysisofthisEEGistodeterminetheirsourceandfunctionalsignicance.Thehippocampaltheta rhythm has been extensively studied in rats and its rhythmicity has recently been shown to befunctionallysignicant,perse.Here, weuserelevantaspectsofthehippocampalliteraturetoprovideperspectiveononeof themost studiedhumanEEGrhythms: frontal-midlinetheta. Wereviewitselectrographic features, localization, prevalence, age distribution, behavioural modulation (particularlyin relation to working memory, spatial navigation, episodic memory, internalised attention andmeditation), relationship to personality, drug interactions, neurochemical relationships, and coherencewith rhythmic activity at other sites. We conclude that FM-theta, like hippocampal theta, appears to playarolein(oratleastoccurduring)processingofmemoryandemotion. Itiscorrelatedwithworkingmemory and/orsustained attention; butthis doesnotentaila roleinfunctionsinceclearbehaviouralcorrelates of hippocampal theta have been demonstrated that are not sensitive to hippocampal damage.FM-theta is increased by anxiolytic drug action and personality-related reductions in anxiety, whereashippocampal theta is decreased by anxiolytic drugs. In animals, frontal theta and hippocampal theta canbe phase-locked or independent, depending on behavioural state. So, the cognitive functions of FM-theta,and their relationship to hippocampal theta, are unclear and denitive evidence for functionalinvolvementincognitiveoremotionalprocessingislacking. Onepossiblesolutiontothisproblemisanalysis of FM-theta in animalsprovided homology can be determined. The issues of sporadicity andlow incidence of FM-theta also need to be addressed in the future. Changes in functional connectivity,indicatedbychangesincoherence, arealsoalargelyuntappedresource. Wesuggestthatthemosthopeful path to assessing the functions of FM-theta will be through the use of drugs, and the variation oftheir effects depending on baseline levels of FM-theta. Finally, we reviewsome theories of theta function.Despite the apparent richness of the current data, we conclude that it is difcult (and may ultimately beimpossible) to formulate a theory that attributes a specic cognitive function to FM-theta. However, thetheoriessharesomegeneralcomputationalassumptionsandtheseshouldbeausefulguidetofuturework and,ultimately,adenitetheoryofthefunctionorfunctionsofFM-theta.2008Elsevier Ltd.Allrightsreserved.* Correspondingauthorat:DepartmentofPsychology, UniversityofOtago, POB56,Dunedin,NewZealand.Tel.: +6434797634;fax:+6434798335.E-mailaddress:NMcN@psy.otago.ac.nz (N.McNaughton).Abbreviations: 5HIAA, 5-hydroxyindoleaceticacid;5-HT, 5-hydroxytryptamine(serotonin);ACC, anteriorcingulatecortex;CNS, centralnervoussystem;EC, entorhinalcortex; EEG, electroencephalography; ERN, error related negativity; ERD, event related de-synchronization; ERP, event related potential; ERS, event related synchronization;FFT,fastFouriertransformation;FM-theta,frontal-midlinetheta;GABA,gamma-aminobutyricacid;iEEG,intracranialelectroencephalography;LORETA,low-resolutionelectromagnetictomographyanalysis;MAO, monoamineoxidase;MEG,magnetoencephalography;MRI, magneticresonanceimaging;REM,rapideyemovement;STAI,SpielbergersStateTrait AnxietyInventory.Contentslistsavailable atScienceDirectProgressinNeurobiologyj our nal homepage: www. el sevi er . com/ l ocat e/ pneur obi o0301-0082/$seefrontmatter 2008ElsevierLtd.All rightsreserved.doi:10.1016/j.pneurobio.2008.09.0052. Electrographicfeatures. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1592.1. Whatistheta?.. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1592.2. WhatisFM-theta? . . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1592.2.1. FM-thetaintheEEG. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1592.2.2. FM-thetainthemagnetoencephalogram:Frontal-mental-theta . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1592.2.3. SignalprocessingandanalysisofFM-theta andtheERPproblem.. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1602.3. WhatisnotFM-theta?. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1602.4. WhatmightbeFM-theta? .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1602.5. Electrographicfeaturesofhippocampaltheta.. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1613. Localization.. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1623.1. LocalizationofFM-theta . .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1623.2. Localizationofhippocampaltheta.. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1634. Prevalenceandage distribution. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1635. Behaviouralmodulation.. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1645.1. EventsthatmodulateFM-theta.. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1645.2. Workingmemory studies... .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1665.3. Spatialnavigationtasks .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1685.4. Episodicmemorytasks .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1685.5. Internalisedattentionandmeditation. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1695.6. PracticalandothertasksthatmodulateFM-theta .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1696. Personalityfactors,drugsandneurochemistry .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1696.1. Personalityandtheta .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1706.2. Druginteractionswithhippocampaltheta .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1706.3. DruginteractionswithFM-theta. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1706.3.1. Anxiolyticdrugs. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1706.3.2. Alcohol andFM-theta.. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1726.3.3. Adrenergicanddopaminergicdrugs.. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1726.3.4. Other psychotropicsandFM-theta . .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1726.4. Neurochemistry. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1736.5. Personality,drugsandneurochemistryconclusions .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1737. Coherencestudies. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1737.1. Human studies:coherencebetweensurfaceEEGanddeepbrainregions . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1747.2. Human studies:coherencebetweenFM-theta andotherbrainregions.. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1747.3. Animalstudies:thetacoherence betweenhippocampusandotherbrainregions. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1758. Discussion. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1758.1. PossibletheoriesofFM-theta . .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1768.1.1. Theorygeneral considerations.. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1768.1.2. Loopselection andcontext encoding . .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1768.1.3. Recursionasgure-groundresolution ofgoal conicts .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1768.1.4. Phase asanindextospatial,episodicandsemanticcontext . .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1778.1.5. Common featuresofthetheories .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1778.1.6. TheoreticalissuesforFM-theta.. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1788.2. DoesFM-thetahavefunctionalsignicance?.. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1788.3. IsFM-thetarelatedtohippocampaltheta?.. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1798.4. WhatisthesourceofFM-theta?. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1798.5. HowcouldFM-thetainteractwithhippocampaltheta? . . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1798.6. SimilaritiesanddifferencesbetweenFM-thetaandhippocampaltheta.. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 1808.7. Mainconclusionsandfuture directions .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. 180References . .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. 1811. Introduction1.1. AfocusonthetarhythmsElectroencephalographic (EEG) recordings from the surface ofthe scalp are replete with rhythmic components. In humans, theEEG rhythms are usually dividedinto different frequencybandsincludingdelta(04 HZ), theta(48 Hz), alpha(812 Hz),beta (1330 Hz) and gamma (40 Hz). However, there are oftendiscrepancies between studies regarding the specic parametersof these bands. The choice of a specic frequency bandandsubdivisions within a band are intended to map bands to distinctbrain functions, and so to map themto physiological andpsychological processes that are specically associated with thatband or sub-band. Amajor task of EEGanalysis is to determine thesource and functional signicance, if any, of the observedrhythms.This reviewfocuses on the rhythmdescribed as frontal-midlinetheta(FM-theta) inthehumanEEG. ThedenitionofFM-theta is consideredbelow. Althoughtheta recordedfromother scalplocationssuchasabovethemedial temporal lobe(GuderianandDuzel, 2005) is startingtobestudiedinmoredetail, thebulkof researchtodatehas focusedonFM-theta.Likewise,althoughtask-relatedalpharhythmhasbeenstudied,task-relatedchanges inthetahavebeenstudiedmoreexten-sively. Forthepurposesof theinitial inclusionof datainthisreview, FM-thetawastakentopotentiallyincludeallrhythmicoscillationsthataremaximal inthe frontal-midline region: F3,Fz,F4;have a frequency in the range 48 Hz or, potentially, somewhathigher;andcanbeobservedeitherintherawEEGorthrougheventrelatedaveraging oranalysis ofpower spectra.D.J.Mitchelletal. / ProgressinNeurobiology86(2008)156185 157AlthoughthemainfocusisonFM-theta, sodened, detailsregarding the properties of other theta rhythms particularlyhippocampal theta in the rodent will be outlined, wherenecessary. In the body of the review our purpose in making suchcomparisonshasbeentoprovideacontextforinterpretationofparticular results obtained with FM-thetaand we have notassumedanyspecicfunctional relationshipbetweenFM-thetaand the rhythms with which we have compared it. However, twomajor questions addressed in the discussion are whether FM-thetaisof functional signicanceandwhetheritisconnectedtothethetarhythm generatedbythehippocampus (Arnolds et al.,1980b; Bland, 1986; Canteroet al., 2003; Ekstromet al., 2005;Green and Arduini, 1954; Halgren et al., 1978; Meador et al., 1991;Raghavachari etal., 2006;Rizzutoetal., 2003;Sainsbury, 1998;Sano et al., 1970; Vanderwolf, 1969). Critically, if differentbehaviours or functional states are associated with the thetarhythmrecordedinhumans versus that recordedinrats thendifferent functional mechanisms are implicated in their generation(OKeefeandBurgess, 1999). Conversely, inasmuchashumansupercial thetaat anysiteishomologoustorat hippocampaltheta,thereisthepossibility ofatransferofthelargeamountofknowledgeaccumulatedregardingrat hippocampal, andtheta,functionstohumantheta. Toassessthis, comparativeresearchneeds to be carefully conducted using paradigms in humans thatcloselyresemblethoseusedwithrats. Amajorpurposeof thisreviewis to generate specic testable hypotheses for futureinvestigations.1.2. FromrhythmtofunctionRhythmic oscillationis likely to occur inexcitatory neuralcircuitsthat arerecursivelyconnectedwhenever thesearenothighlydamped. Epilepticseizuresareanextremeformof suchnon-damped activity. Such intrinsic oscillations, like the hummingof a string in the wind, are a consequence of a simple owof energythrough an elastic systemand, since they result fromhugenumbers of neurons ring synchronously, reect a lack of detailedcodingofinformation. Assuch theserhythms arelikelyeither tolackfunction or tobesignsof dysfunction.Functionally signicant rhythms can occur when a specicexternal pacemaker forces signicant numbers of neurons (whetherrecursively connected or not) to re and become silent atapproximately the same time as each other. The pacemakerdetermineswhenneuronsmayrebut, providedonlyamodestproportion of neurons are active at any one time, large amounts ofinformationcanbeencodedbywhichneuronsinfactreattheappointed time. The hippocampal theta rhythm (which is in theregion of 46 Hz in animals such as rabbits, cats and dogs but occursintherange514 Hzinfreelymovingrats) is controlledbyapacemaker in the medial septum(Brucke et al., 1959; Gogola k et al.,1967, 1968;Stumpfetal., 1962). Thissendsaninhibitoryphasicsignal tothehippocampus(Leung, 1998) andneuronsreceivingexcitatory input re predominantly when the inhibition is released(Smytheet al., 1992). Informationmaybeencodednot onlybywhichcells re but also by the precise phase relation of their ring to theongoing rhythm(OKeefe and Recce, 1993). Similarly, cortical alphaisconsideredtobegenerated, onoccasion, byinputfromothercortical andthalamicnetworks(Klimesch, 1999). Itislikelythatcontrol of phasic ring by pacemakers is important for the control ofprocessingbytheloops that pass informationrecursivelyfromneocortextothehippocampus (Miller, 1989, 1991) or aroundrelatedcircuits (Buzsa ki, 2006; Parmeggiani et al., 1971).It is possible that, at times, other external pacemakers than theseptumcontrol hippocampal theta; andthat thehippocampuscould showtheta as a formof auto-oscillation. However, at present,webelievethereisnoevidencethat thisoccursunder normalphysiologicalcircumstances infreelymoving animals.Until recently, it was not knownwhether thesynchronousrhythmic activity giving rise to any EEG rhythmhad directfunctional signicance or was just a correlated epiphenomenon offunctional activity.However, it hasrecentlybeen shown,usingabrainby-pass techniquethatrestoringhippocampal rhythmi-city, withoutanyrestorationofthenormalspatialandtemporalpatterns of individual cell ring, can restore psychological functionthat is lost when rhythmicity is blocked (McNaughton et al., 2006).This nding with hippocampal theta suggests that other rhythmsmay have similar functional signicance. However, as we shall see,what is known about the occurrence and functional signicance ofhippocampal theta implies that there are major problems in usingcorrelational approaches to determine the underlying functions ofanybrain rhythmicity.1.3. AhippocampalperspectiveInthisreview, weassesstheextent towhichFM-thetahasrelated properties to those of hippocampal theta and thetarecordedfromothercortical brainregions. Themainreasonfortaking a hippocampal perspective is that, at the general level, thehuge literature on hippocampal theta provides a strong frameworkfor asking detailed questions about FM-theta without any need toassumeidentityorhomologybetweenthetwo. That is, knownrelationsorlackthereof betweenhippocampal theta, behaviourand function can provide a basis for judging the value of equivalentevidence in relation to FM-theta. However, at the specic level, wealso need to determine to what extent hippocampal theta and FM-theta might be directly related. We will argue that each may occurindependently while also, on occasion, becoming locked togetherincoherentoscillations.We will start by dening FM-theta, and listing similarphenomena that are clearly not FM-theta, and then considerphenomenathatareindeterminate.Wewillthenconcentrate onclear FM-theta and outline itsfeatures; electrographiccharacter-istics; possible sources; prevalence; behavioural associations;neurobiological associations; and drug interactions. We will also,wherepossible, comparetheseat ageneral level withwhat isknown of hippocampal thetawhich can often illustrate potentialpitfallsininterpretation ofthe humanEEG data.1.4. PragmatismversustheoryInthebulkofthisreviewwehavetakenalargelypragmaticapproach that avoids a strongly theoretical focus. There are threereasonsforthis. First, webelievethatabalancedand, asfaraspossible, comprehensivepresentationof thedataas theynowstand will more readily lead to a proper theory of the function orfunctions of theta if it is not ltered by current theoreticalpreconceptions. Second, as we discuss inSection 8.1, currenttheories appear to have incommonthe viewthat theta actsgenerally to control processing in neural circuits and so itsapparentcognitivefunctionatanypointintimewilldependonwhichcircuitsitiscontrollingratherthanbeinginherenttoitsrhythmicity. Thirdly, the bulk of the data are necessarilycorrelationaland here the hippocampus provides us with auseful yardstick. Despite seven decades of hard work on rabbits,rats, mice, gerbils, guinea pigs, sheep, cats, dogs, Old Worldmonkeys, chimpanzees and humans by outstanding colleagues, todate, there is still no agreed termthat would unequivocallydescribebehaviouralcorrelate(s)ofhippocampalthetarhythms(Buzsa ki, 2006, p. 21). For all these reasons we will leavetheoretical considerationsto thenal section.D.J.Mitchelletal. / ProgressinNeurobiology86(2008)156185 1582. Electrographic features2.1. Whatistheta?The theta frequency in the human EEG was initially dened as47 Hz (Walter andDovey, 1944) simplytoprovideaninter-mediate frequency band between delta and alpha. The name thetawas given with reference to the thalamus where the rhythm wasthought toarise. Later, thetafrequencywasdenedas48 Hz(IFSECN, 1974). Ifsuchbandsaredenedbyindependentneuralprocesses, it is clearly open to question whether they are exclusive.Can theta, functionally dened, extend into the alpha range, forexample?FM-theta in normal, awake, adults has been the focus ofresearchbecauseitsamplitudeanddurationareexceptional. Atother sites there is, by and large, only a small amount of power attheta frequencies and no clear bursts in individual traces(Niedermeyer, 1999). Important distinctions canalsobemadebetweentonicandphasictypesof theta(Klimesch, 1999). It issuggestedthat phasic andtonic types of theta have differentfunctions and generators. Both phasic and tonic forms of theta havebeen correlated with a number of different behaviours. Inparticular phasic theta occurs in response to some event oractivity and has a discrete temporal and topographical appearanceover frontal regions. In contrast tonic theta tends to coincide withmore global and stable characteristics or phenomena and to have amore diffuse topography. Hippocampal theta was so calledbecause its initial discovery was in animals in which hippocampalrhythmicitywas inthe conventional humanEEGtheta range.Hippocampal theta is now dened less by its frequency than bythe fact that its sources and sinks result frominput fromthe medialseptum (and produce rhythmic ring of hippocampal neurones) inall species. This homology has resulted in such hippocampalrhythmicity being termed theta in all species independent of theprecise frequency band observed. As we noted above, thefrequencyofhippocampalthetacanbeashighas14 Hzintherat. Hippocampal theta rhythm in the rat is consistently generatedprior to certain movements (Morris and Hagan, 1983; Vanderwolf,1969) but also occurs in the complete absence of movement underconditions of high arousal (Sainsbury, 1998). In both cases it can beviewed as a sign of preparation for movement but in the latter caseitisapreparation that doesnotinfact resultinaction.The hippocampus has been postulated to contribute tocognitivefunctionssuchasspatialmapping(OKeefeandNadel,1978) and relational memory (Cohen and Eichenbaum, 1993) andto emotional functions (Papez, 1937), particularly behaviouralinhibition (Gray, 1982; Gray and McNaughton, 2000). Allcomponents of the hippocampal formation, including the entorh-inal and posterior cingulate cortex as well as hippocampus proper(Gray and McNaughton, 2000; Leung and Borst, 1987), show thetaactivity.Activity in the medial prefrontal cortex can be modulated by thehippocampusdependingonthebehavioural stateoftheanimal(Jones and Wilson, 2005). This raises the issues of howfarhippocampal theta may contribute to frontal theta and, given theknown extensive variation in the frequency of the former, what isthe frequency bandofthe latter.2.2. WhatisFM-theta?2.2.1. FM-thetaintheEEGThetermFM-thetawascoinedbyIshiharaandYoshi(1972).They dened it as a distinct frontal-midline theta rhythmat 67 c/sand 3060 mV lasting seconds during mental tasks such ascontinuous arithmetic addition. Table 2 presents informationfromstudiesthateitherprovidedFM-thetasignal properties, ordescriptive illustrations (an example of FM-theta is given in Fig. 1).These papers suggest that FM-theta is almost always around 6 Hz,thoughitcanrangefrom5to7.5 Hz(betweenindividuals). Theamplitude is around 5075 mV (and often higher than the averagebackground amplitude). The waveformis basically sinusoidal,superimposedontheprevailingbackground. Withinindividuals,the frequency appears consistent within bursts and also betweenbursts within illustrated cases. Between individuals there aredifferences, and some do not demonstrate FM-theta at all.Typically, FM-theta appears indiscrete bursts that last a fewseconds but can range from 1 to >10 s. It tends to wax and wane(Mizukietal., 1980)andtheredoesnotappeartobeanytime-lockedrelationshipbetweentheinitiationof thetaskandtheappearanceof FM-theta(Gevinset al., 1997). For example, theamount of FM-theta rangedfromabout 20to110 s during acontinuous Uchida-Kraepelin task lasting 5 m(Mizuki et al., 1980).(TheUchida-Kraepelintaskinvolvescontinuousarithmeticaddi-tionofsuccessivepairsofnumberswhere thesecondnumber ofthe previous pair is the rst of the current pair.) FM-theta,therefore, can be described as a phasic, task-related modulation ofthebackgroundEEG. However, somestudieshaveclearlyshownphase- andtime-lockedcomponents during behavioural tasks.Where there is sufcient electrode cover to estimate the eld, it isalways fronto-central, with amaximum at,or just anterior toFz.2.2.2. FM-thetainthemagnetoencephalogram:Frontal-mental-thetaMagnetoencephalographic(MEG)recordingscanalsoidentifyfrontal-midline theta rhythms and provide complimentary data toEEG studies. MEG provides an insight into the tangential currentsgeneratedduringencephalographicactivityandislesscontami-nated by volume conducted currents through the layeredstructuresabovethecortex.Assuch, itisbettersuitedtosourceanalysis.Sasakietal.(1994b)coinedthetermfrontalmentaltheta toaccount for theMEGphenomenonof thetawaveforms duringFig. 1. Illustration of frontal-midline theta recorded during the performance of a continuous N-back working memory task. Two seconds into the EEGtrace 6.57 Hz theta canclearly beobservedat F3,Fz,andF4with anamplitudeof 5075 mV.Thethetaactivityhasapeak atFzandtendstowax andwaneovertime.D.J.Mitchelletal. / ProgressinNeurobiology86(2008)156185 159calculation and musical imagining and have subsequently studiedMEGcomponents of FM-theta (Sasaki et al., 1996b,c). These studiesdemonstrated that FM-theta and frontal-mental theta haveidentical frequencies, and similar waveforms (Sasaki et al.,1996c), althoughslightlydifferingdistributions. Frontal-mentaltheta appears more broadly distributed in the frontal lobes whencompared to FM-theta. But this is likely to reect the differences inthe features recorded by the two techniquesin studies where FM-theta has been synchronously recorded using scalp EEGand MEGitisreasonabletoassumethat frontal-midlinethetaandfrontal-mental theta represent different projections of the same phenom-enon. Iramina et al. (1996), Ishii et al. (1999) and Asada et al. (1999)have also employed synchronous EEG and MEG to study FM-theta(seeunderlocalization ofFM-theta fordetails).For the purposes of the remainder of the review, we willprovisionallyassumethatfrontal-mentalthetarecordedbyMEGandFM-thetaintheEEGrepresentthesameunderlyingneuralphenomena. However, further detailed research is clearly requiredtodetermine whether thisisthecaseor not.2.2.3. Signal processing and analysis of FM-theta and the ERP problemAnumberof techniquesareavailableforanalysingtheEEGsignal including averaging and event related potential (ERP)analysis, event relateddesynchronization/synchronizationERD/ERS (Pfurtscheller and Lopes da Silva, 1999), fast Fouriertransformation (FFT) (Thakor and Tong, 2004), and waveletanalysis(Basaretal., 2001a). Thesetechniquescaptureinforma-tion pertaining to time, power, and the frequency domain. Signalprocessing techniques such as FFT, ERD/ERS analysis, and waveletanalysisareauseful methodofdatacompressionandcanrevealfeaturesof FM-thetaunderappropriateconditions. FFTanalysisreveals the prominent frequencies within an EEG segmenthoweveritdoesnotprovideadetailedaccountofthetemporalcomponents of the EEG. In contrast, ERD/ERS (Pfurtscheller, 1992;Pfurtscheller and Aranibar, 1977, 1978; Pfurtscheller and Lopes daSilva, 1999) and wavelet analysis can help to determine changes inspectral powerwithabettertemporal resolution. Assuchtheyprovide powerful tools for investigating events that modulaterhythms such as FM-theta (e.g. behaviours and drugs). If a subjectperforms a task known to elicit FM-theta, and this is conrmed byvisual analysis, thenappropriatesignal processingof datawillreveal a peak around 57 Hz in frontal-midline electrodes(Harmony etal., 1999; Pellouchoud et al.,1999).However,itshouldbenotedthatthereverseneednotbethecase. The Fourier transform, on which both simple power spectraand wavelet analyses are based, will convert any data series into aset of sine wave components. The presence of power at a particularfrequency in a transform, then, does not provide any evidence thattheunderlyingsignalisrhythmicinthesenseofrepeatingovermultiplecycles. Powerinthethetarangecould, forexample, begenerated by evoked potentials. Strictly, then, to interpretaveraged power changes as evidence for changes in theta rhythmeithersinusoidal rhythmicityshouldbedemonstratedinindivi-dualrawtracesorviaautocorrelationofrawtraces.Thiskindofcheck has often not been undertaken in the studies reviewed here.There are a number of studies and theories that emphasise theimportanceofthephaseofEEGoscillationsinworkingmemory(Givens, 1996;JensenandLisman, 1998), retrieval, andencodingprocesses (Hasselmo et al., 2002; Rizzuto et al., 2006). In contrast tosignal processingtechniques, averagingof therawEEGis aneffectivemethod to tease out low amplitude event related potentials (ERPs)that are time- and phase-locked to an event while eliminating othersignals. Acritical questionarisingfromthis lineof researchis towhatextent the averagedERPis composedof discrete evokedpotentials ofa similar conformationandtowhat extent it results fromphase resetof ongoing rhythmicityand to what extent it contributes to powerinsignal processing (BastiaansenandHagoort, 2003). Analysis of theERP does reveal that components of this waveform are within thetheta frequency range (Brankack et al., 1996; Bruneau et al., 1993;YordanovaandKolev, 1996).Therefore, averagedERPsatfrontal-midlinesites maybea formof FM-theta, or functionallyhomologousto it (Bruneau et al., 1993; Burgess and Gruzelier, 1997).Furthermore, aninverserelationshipbetweenfrontal ERPs andpre-stimulus theta components has beendemonstrated(Basar et al.,2001b; Yordanova and Kolev, 1997a, 1998). With the possibility of aconsistentphase-reset, itisdifculttodisentanglewhetherERPssuch as the N200-P300 complex are the result of phase-resetting oftheongoingEEGortheevocationof transient EEGevents, oracombinationof thetwo. Tosomeextent this issuecanbesettledwithsimple averaged waveforms. Reset of ongoing theta in thehippocampus, for example, producesaclear dampedoscillationwith multiple peaks and troughs and a generally consistent peaktroughandtroughpeakinterval(Givens,1996).Conversely,onasingle sweep basis, an electrically evoked potential produces onlyone or two negative and positive components (Brankack et al., 1996)and, whenaveraged, theseresult inwaveformsthat areunlikedampedoscillationinterms of boththe limitednumber of peaks andtroughs and the marked variation in peaktrough and troughpeakintervals.However, averaging the human scalp EEGmay not be aneffective method to detect phase-reset of a potential hippocampalhomologue. In contrast, techniques involving wavelet analysis andphase-locking statistics may be a more effective approach todetermining phase-resetinhumans. Currentresearchhasshownthat phase-reset does contribute to event related potentials,particularly to early components such as the P1-N1 complex withpeaks at 100 and 150 ms, respectively (Hanslmayr et al., 2007;Klimeschetal., 2004;Makeigetal., 2002), andtheerrorrelatednegativity(Luuet al., 2004;Yeungetal., 2007).2.3. WhatisnotFM-theta?For the purpose of this review we shall exclude theta rhythmsassociated with pathology. These include: epilepsy (Ciganek, 1961;Mokranet al., 1971; WestmorelandandKlass, 1986); cerebraldysfunctioninchildren(WhiteandTharp,1974);Rettsyndrome(Niedermeyer et al., 1997); duringpost-anoxiccoma(Berkhoffet al., 2000); attention decit hyperactivity disorder (Lazzaro et al.,1998; Mann et al., 1992); high pressure nervous syndrome (Okudaet al., 1988); and 50 Hz electrical stimulationof the anteriorcingulate cortex(Talairach etal., 1973).Some theta rhythms are clearly not fronto-central and so mustbe treated as separate fromFM-thetae.g. intracranial thetarecorded fromparietal and occipital sites during the Sternberg task(Raghavachari et al., 2006). However, theta rhythms recorded fromotherlocationsonthescalpandintra-cranially, thatmaybeofrelevance to FM-theta, will be discussed belowwhere appropriate.We shall exclude some of the normal, sleep related thetarhythms: theposteriorlowamplitudethetathat occursduringdrowsiness; the diffuse medium voltage theta that increases withincreasingdrowsiness(Janatietal., 1986);the26 Hzsawtoothwaves apparent fronto-centrally during REMsleep; and theapparent FM-theta that has been reported during NREM and stage1 of sleep (Hayashi et al., 1987). These types of theta are generallydifferent inboth formand distribution toFM-theta.2.4. WhatmightbeFM-theta?There are some rhythms and frequency components that are in,orclosetothethetaband,thatoccurinnormalparticipantsandD.J.Mitchelletal. / ProgressinNeurobiology86(2008)156185 160that mayor maynot beFM-theta. Someof theserhythmsareclearly fronto-central. Bursts (12 s increasing to 10 s-severalminutes) of fronto-central theta rhythms (57 Hz) have beendemonstrated during meditation (Banquet, 1973; Hebert andLehmann, 1977). Hayashietal. (1987)identiedfrontal-midlinetheta during specic sleep stages (mainly stage 1 and REM sleep)that were often associated with inner experiences with a distinctcontent. Takahashi et al. (1997) found similar characteristicsbetweentheFM-thetarecordedduringamental taskandthatassociated with drowsiness. Furthermore, participants whodemonstratedFM-thetaassociatedwithdrowsinessalsodemon-strated FM-theta while performing a task. This appears to create aparadox that FM-theta occurs in both a state of alertness or xedconcentrationanddrowsiness. However, itmaybecausedbyaninhibitory mechanism that blocks out information when going tosleeporwhentryingtoretainsomethinginworkingmemoryorwhenfocussingattention inmeditation.Conversely, these rhythms while appearing onthe surface tobesimilar, may yet have quite distinct mechanisms. Therefore, withthese rhythms, the issues are howfar they share the psychologicalfunctional correlates and the electrographic and topographicfeaturesofconventionalFM-theta. Thekeywillbeindesigningexperiments andmethodologies that attempt todifferentiatewith more precision the electrographic features of differentpsychological processes such as working memory and meditativestates.Forexample, Shinomiyaetal. (1994)conductedastudythatrevealed two distinct types of frontal-midline theta rhythm. Type 1FM-thetawasmoreprominentinhealthysubjects, hadahigherfrequency, regularshape, lowervoltage, andashorterdurationthanType2FM-thetathatwasmorecommoninpatientswithepilepsy. Furthermore, Type1tendedtospreadbilaterallyandanterior toFzwhiletheType2thetawas spreadtoposteriorregions. Similarly, Gevins and Smith (1999) used a neural networkalgorithm to differentiate between a specic frontal-theta relatedtotaskdemandsandadiffusethetaassociatedwithfatigueandhangover. In both these cases, then, quite careful parameteranalysis is required to distinguish what appear to be functionallydistinctbutsupercially quite similar rhythms.One explanation of their being two types of FM-theta is thatquitedistinctcircuitsareinvolvedineachcaseand, potentially,non-overlappingneural sourcesof therecordedrhythms. How-ever, with rodent hippocampal theta, two pharmacologicallydistincttypeshavebeenidentied(Bland, 1986)acholinergic-dependent type of theta occurring during immobility and one thatisnotcholinergicallydependentthatoccurswhentheanimalismoving. These appear to involve the same frequency controlcircuitsand, inessence, canbeenseenasthesamefundamentaltheta that is gated by different modulatory inputs at differenttimes(Gray andMcNaughton, 2000).Thetwotypes ofFM-theta,then, couldinvolvethesamecorecircuits, but releasedunderdifferentconditionswiththedifferingreleaseconditionsaffect-ing which other brainregionsarerecruited.Some theta rhythms have been attributed to emotional tension.Mundy-Castle (1951, 1957) reported anemotionally elicited 56 Hzrhythmat various locations onthescalp. Sasaki et al. (1996a)reported fronto-parietal MEG theta rhythms elicited by emotionalthoughts. These were of slightly higher frequency (6.3 Hz) than theMEG FM-theta reported in that study. With these rhythms it is notclear how far their frequency, topography or functional correlatescan be identied with or distinguished from FM-theta.Other possibly relevant rhythms are also found ininfants andchildren. Generalised, high amplitude theta is a well known featureof theEEGof drowsyinfants. Maulsby(1971)demonstratedanemotionallyelicitedposteriorrhythmof4 Hz, ina9-montholdchildoldand Orekhovaet al.(1999) demonstrated afrontal3.64.8 Hz rhythmin 811 months old children during anticipation in apeekaboo game. Orekhova et al. (1999) also demonstrated a 5.26 Hz temporal rhythmthat they suggestedmay have hadanemotional substrate. Kugler and Laub (1971) reporteda 4 Hzposterior rhythms inyoungchildren(>6months) and56 Hzfrontal rhythmsinolderchildren(upto6years)duringpuppetshows or similar experiences. Theta rhythmcanoccur inthetemporal and parietal regions in response to some formof externalstimulation that causes the infant to orientate and direct attention(Kugler, 1973) anddiminishesafter afewseconds(KuglerandLaub, 1971). Kugler and Laub (1971) also noted that the theta tendstobemoreprominentinposteriorlocationsinyoungerchildrenandhas amoreanterior or central locationlater on. Alsothefrequency is typically 4 Hz in younger children and 56 Hz in olderchildren. Afrontal-midline67 Hzthetarhythmisarecognisedfeature in the EEG of awake children (Niedermeyer, 1987). Theseresults raise the question of howfar adult FM-theta and itsvariations across subjects represent maturational restriction of theoccurrence of theta and how far the theta components in childrencanbeseenasdistinct fromadult FM-thetaandlikethethetacomponents thatareseen inadultpathology.In summary, there are a variety of psychological states that areassociatedwiththetarecordedat frontal electrodesites. Whenattempting to distinguish FM-theta from other EEG features, it isimportanttonotethatsomeabnormalthetarhythmsandsomenormal rhythms identied in children are almost indistinguishablefromFM-theta in terms of the waveformcharacteristics andtopographyoutlinedinSection2.2.1. Wewillreturntosomeofthese rhythms inthediscussion.2.5. Electrographicfeaturesofhippocampal thetaThe electrographic features of FM-theta describedabove liewithinthe domainof expectedfeatures of hippocampal theta.Hippocampal thetahasbeendifcult toisolateinhumans. It isunlikelythat hippocampal theta wouldbeprojecteddirectly toscalpelectrodesbyvolumeconduction, andthereforewearegenerallyrestricted to the rare occasions where electrodes are placed directlyintothehumanhippocampusinsubjectswithsevereunderlyingpathology. Thoseinvestigationshaveindicatedarangeofhumanhippocampalthetafrequencies.Sanoetal. (1970)applied100 Hzstimulation to the posterior hypothalamus in patients undergoingsurgery and often observed theta (67 Hz in the gure presented asan example) in electrodes placed in the hippocampus. Halgren et al.(1978) identiedabehaviour modulated56 Hzrhythminthehippocampus of a subject undergoing monitoring for epilepsysurgery. This 56 Hz rhythm desynchronised during certainbehaviours, but the authors suspectedthat it was abnormal. Arnoldset al. (1980a) identied a behaviour-modulated 34 Hz hippocam-palrhythm. Isokawa-Akessonetal. (1987)recordedextracellularsingle-unitactivitiesfrom theanteriorhippocampusthroughneplatinummicroelectrodes. They investigated 23 subjects, andanalysedringratesusingautocorrelationanalysisandinferredclear rhythmic components in the range 6.717 Hz.More recently, MEGandsignal processingtechniques havebeenusedinattempts tostudyhippocampal theta innormalsubjects(ItshouldbenotedthatthecapacityofMEGtodosoiscontroversial.). Tesche (1997) useda full headMEGarray onnormal subjects during mental arithmetic andpassive pictureviewingandidentiedcomplexactivity, attributedtothehippo-campus, withspectral features withcomponents below12 Hz,includingsometaskdependent peaks. TescheandKarhu(2000) havesince reporteda task-specic 7 Hz theta component insignalssuggested to originate in the hippocampi of normal subjects.D.J.Mitchelletal. / ProgressinNeurobiology86(2008)156185 161Theta, when recorded directly fromthe human hippocampus, isgenerallyidentiedusing signal processing andis not alwaysclearly apparent in the rawEEG(but see, e.g. Sano et al., 1970). Thisisincontrasttohippocampalthetainnon-primates, particularlyrodents, usually used to study hippocampal theta. This may be dueto underlying pathology, electrode location, activation procedures,conformationofthehippocampal celllayersor aconsequence ofthe broadrange of ongoing activity that occupies the humanhippocampus. Preciseelectrodelocationaffectstheamplitudeofthetamarkedlyinrodents(Robinson,1980)andisclearlyoneofthe most important variables toassess infutureexperiments.However, it should be noted that theta appears as difcult to recordin other primates as in humans (Crowne et al., 1972; Stewart andFox, 1991) and also canrequire the use of signal processingtechniques todemonstrate it(Tsujimoto et al.,2006).3. Localization3.1. LocalizationofFM-thetaWhile many studies of FM-theta use a small number ofelectrodes concentratedaroundthe regionof interest, severalinvestigatorshaveattemptedtolocalisethesourceof FM-thetawithhighresolutionsurfaceEEGrecordingsorwithpairedEEG/MEG recordings. There are likely to be multiple generators of thetarhythm recorded in the human EEG (Raghavachari et al., 2006). Todate several brain regions have been implicated as potentialsourcesof FM-theta.FM-theta is generally maximal at, or just anterior to, Fz withlittleposterior propagation(Yamaguchi et al., 1990b); but itsmaximumamplitude can shift within 3 cmaround Fz, even withina single subject (Ishihara et al., 1981). The localised distribution ofFM-theta power indicates that it probably originates fromasource on or near the frontal midline. It should be remembered,however, that sourceanalysis basedonscalpEEGis difcultbecauseoftheinverseproblem(Koles, 1998). Forexample, theN100-P200complexfor lateauditoryevokedpotentialshas amidline maximumthat was initially thought to originate fromthefrontal lobesor widespread, diffusesources(e.g. Pictonet al.,1974) but it is now thought to be localised to a group of bilateralsourcesintheauditory cortices(Scherg et al.,1989; Scherg andVon Cramon, 1985).Sasaki et al. (1994b) were the rst to look for a source for FM-theta using synchronisedMEG/EEGandsuggestedthat it wasgeneratedbymultiplebilateral sourceswithinthefrontal lobesandtheysuggest that themidlinemaximumis duetovectorsummation(Sasaki et al., 1996c). The MEGstudies by Sasaki(Sasaki et al., 1994b, 1996c) havesometimesdemonstratedanasymmetry in the MEG maxima, with the right fronto-central areahavingaclearlyhigher signal.Inouye et al. (1994b) noted a lateral to medial potential owofFM-theta in areas in front of Fz with a medial to lateral potentialowinareasbehindFzduringtheperformanceoftheUchida-Kraepelintask. ItwasalsosuggestedthatFM-thetamovesinacircularclockwisemotionintheleftfrontal regionsandantic-lockwise in right frontal regions. Iramina et al. (1996) suggested asingle midline dipolesource, but questioned the validity ofthatmodel. Ishii et al. (1999) usedMEGtechniquesandproposedsources in the bilateral medial prefrontal cortices, includinganteriorcingulatecortex(ACC). Theypointedout that if largeareas of frontal cortex were involved in generating FM-theta thenpointsourcedipoleanalysismaynotbeappropriate. TheyalsosuggestedthatthethalamusmightbedrivingFM-theta. Asadaetal. (1999)usedEEGcombinedwithMEGandproposedtwosourcesintheregionsbothoftheprefrontal-medialsupercialcortex and ACC. They suggested that these regions werealternatelyactivatedduringoneFM-thetacycle. Gevins et al.(1997) usedhighspatial resolutionEEGinconjunctionwithMagnetic Resonance Imaging (MRI) and suggested a medialprefrontal source in the region of the anterior cingulate but notedthat because of the anatomy of this structure it was not possibletodetermineif thiswasorwasnotbilateral. Similarly, Ontonet al.(2005), using a dipole source model, localisedFM-theta totheapproximateregionofthedorsalACC.Sausengetal.(2007)used low-resolution electromagnetic tomography analysis (LOR-ETA) tostudythesourceof FM-thetaduringamotorworkingmemory task. The LORETA revealed the source of FM-theta duringthistasktobetheanterior cingulategyrusandthecingulatemotor area. Finally, Tsujimoto et al. (2006) attempted to localiseFM-thetainmonkeyswhileperformingataskwherebyaleverhadtobeliftedforacertainperiodbeforearewardwasgiven.Releasing the lever resulted ina waiting period. Increased thetalocated in the rostral ACC and prefrontal cortex occurred prior toand following hand-movements. Rewarding the hand movementresultedinasecondarygaininthetaamplitude. Althoughthisndingisperhapsconsistentwiththehumandataitisunclearwhether monkeyandhumanfrontal-midlinethetaarehomo-logous.All of theseanalyses suggest afrontal lobe/ACCsourceorsources for FM-theta. This is consistent with the ACC beingconnected to a number of different brain regions and involved in arangeofcognitiveandemotionalfunctionssuchasmotivation,decisionmaking, processing information, andattention(Devinskyet al., 1995; Pardo et al., 1990; Posner and Petersen, 1990; Wangetal., 2005). However, thisisnot entirelyconsistentwiththerecentndingsofKahanaetal. (1999). Theyidentiedclear48 Hz rhythms from subdural electrode arrays recording from thesupercialcortexinthreesubjectsduringvirtualmazenaviga-tion. ThistypeoftaskcangeneratesupercialFM-theta(videogames and a simulated driving task have been used to elicit FM-theta). However, they had no surface electrodes and so therelation oftheir recordings to conventional FM-theta is unclear.They foundtheta fromrecordings made by electrode contacts overbroadareasofcortexincludingthetemporal lobe, theparietallobe, the inferior frontal gyrus andthe central sulcus. Thetaappeared in discrete bursts of 12 s and occurred more frequentlyin more complex mazes and during recall trials. Given the natureof intracranial EEG(a very high amplitude signal recorded directlyfromthecortexwithnointerveninglayers), itislikelythattherecorded rhythms originated from sources near to the recordingelectrodes, and are probably not attributable to volume conduc-tion. FM-theta, then, mayhavearegularsupercial locationatfrontal sites not because of a frontal generator but because of thewaythatmultiplesubduralgeneratorscombinetoproducethesupercial eld.Cortical thetacouldbelinkedtothetageneratedwithinthehippocampus. Correlations intheoccurrenceof thetabetweenhippocampal and neocortical sites have been noted during a virtualnavigationtask(Ekstromet al., 2005). Furthermore, Fell et al.(2003) found 119 Hz coherence between hippocampal andadjacent rhinal cortex during a declarative memory task. However,otherstudieshavenotdemonstratedaclearlinkbetweenthetarecordedinthe hippocampus andcortical regions inhumans(Canteroet al.,2003;Raghavacharietal., 2006)alsosee(Kahanaetal., 2001).Inrats, frontalcortexandhippocampusoftenshowindependent thetaactivityof different frequencies but, duringexploratory activity, can showcoherence with frontal areasbecominglockedtothehippocampus(YoungandMcNaughton,2008). This observed behaviour-dependence could account for thevariationinndings inthehumanstudies.D.J.Mitchelletal. / ProgressinNeurobiology86(2008)156185 1623.2. LocalizationofhippocampalthetaHuman hippocampal theta has not been denitively localised tosourceswithinthehippocampus.Halgrenetal.(1978)identiedtheta fromall electrode contacts inthe hippocampus of theirsubject and it was most prominent in the contacts in the posteriorhippocampal gyrus. Arnolds et al. (1980a) identieda 34 Hzrhythm in the area of the pes hippocampi. Isokawa-Akesson et al.(1987) identieda clear rhythmof 6.717 Hz inthe anteriorhippocampus. Sano et al. (1970) identied a 5.56.0 Hz in the righthippocampus. However, current source density analysis of the typerequiredtodeterminesourcelocationshasnotbeencarriedout(and forethical reasons isunlikely tobecarried out).In animal studies, hippocampal theta is usually recorded fromlocal elds intheCA1anddentatelayers of thehippocampalformation. ThetarhythmhasalsobeenrecordedfromareaCA3(Villarreal et al., 2007), perirhinal cortex (Bilkey and Heinemann,1999), the cingulate cortex (Colom et al., 1988; Talk et al., 2004),entorhinal cortex (Dickson et al., 1995, 1994), hypothalamus,cortical structures, and amygdala (Seidenbecher et al., 2003).Furthermore, rathippocampalthetaisstronglylinkedtophase-locked theta rhythms in the posterior cortex (Colom et al., 1988),cingulate cortex (Feenstra and Holsheimer, 1979), and cansometimesbelinkedtophasicsinglecellringinfrontalcortex(Siapas et al., 2005). Furthermore, thehippocampus receives anumber of inputsfromthehypothalamusandmedial septum/diagonal bandcomplexthat areinvolvedinthemodulationoftheta(Kirk,1998).The lack of reports of strong frontal theta rhythm(as opposed tophasiccellularactivity) inrodentssuggeststhathuman, super-cially recorded, FM-theta is not likely to be related tohippocampal theta. However, it should be remembered that boththecingulatecortex(Colomet al., 1988; Talket al., 2004) andentorhinal cortex (Dickson et al., 1994) can generate theta and thatthehippocampus has strongconnections tothefrontal cortex(Miller, 1991). FM-thetacould, then, beareectionofrhythmichippocampal cellular activity that does not produce a superciallyrecordable hippocampal rhythm but does entrain frontal cells andproduce, atleastsporadically, rhythmicactivitythatiscoherentwith hippocampal activityand there is preliminary directevidencethat thisisthecase(YoungandMcNaughton, 2008).Theta recorded above temporal cortex could, likewise, result fromentorhinal theta that is not accompanied by a recordablehippocampal rhythm. Thiswouldbesimilartotheentrainmentof hippocampal theta frequency by septal pacemaker cells. Indeed,and more generally, it has been argued that the function of thetaactivity(inpopulationsofneurons)istointegrateorcoordinateactivityacrossadistributedsystemthat mayhaveelementsinneocortex,anddiencephalonaswellasthehippocampus(Bland,1986; Blandand Oddie,2001;Kirk andMackay, 2003).It is also, of course, possible that frontal rhythmic activity couldoccurcompletelyindependentlyof thehippocampus. Certainly,where FM-theta is a different frequency fromtheta activityconcurrently recorded from other sites then oneor both of thesemust be independent of the hippocampusactivitythroughoutwhich isgenerally coherent.4. Prevalence and age distributionSomeinvestigators recruit or analysedataonlyfromthosesubjects that demonstrate reliable FM-theta. For example,restricting analysis to Fz and cases where FM-theta had (1)rhythmical-sinusoidalconguration;(2)markedlyhigherampli-tude as compared to background activity, and (3) durationexceeding1 s(Mizukietal.,1980, p.346).Furthermore,reportson the presence of FM-theta may differ depending on whatparticipants are required to do during the experiment. In fact, notall people display this phenomenon in scalp recorded EEGs(Table1)andthereappearstobeanagedistributionwithFM-thetabeingmost commoninyoungadults, withtheincidencedecreasing signicantly after about 30years of age. Althoughreachingapeakinyoungadults, increasesin45 Hzthetaoverfrontal and prefrontal regions were noted during a peekaboo gamethat involved both internal (or anticipatory) and external attentionin 78-month-old infants (Stroganova et al., 1998). Activity at 56 Hz was notedtoincreaseonlyduringtheinternal attentioncomponent. Attheotherextreme, inanoldersampleaged 69years, FM-theta did not increase with working memory taskdemands (McEvoy et al., 2001). Why this age-related peak occurs isunclear, but insomewaysthismimicstheprevalenceandagedistribution of scalp recorded mu rhythm (a rhythm in the alpharange that is maximal over somatosensory cortex and showssuppression during contralateral movement) which is identied inabout8%ofthepopulationwithapeakofabout13.8%in1020years olds (Niedermeyer, 1987). The opposite pattern is noted withalpha rhythm. It is interesting to note, however, that when MEG orTable 1PrevalenceandagedistributionofFM-thetaInvestigators Agerange Sample size FM-thetapresence Numberofsubjects displayingFM-thetaAsadaetal.(1985) 1923years 191 Yes 71 37%Cummins andFinnigan(2007) 1827and6080years14npergroup Greaterincreasewith taskdifcultyinyoungergroup IshiharaandYoshi (1972) 18.3years(mean) 115 Yes 60 52%McEvoyetal.(2001) 21 years 10 Increasedwithtask difculty 47 years 10 Increasedwithtask difculty 69 years 10 Decreasedwithtask difculty Mizukietal.(1980) 2026years 30 Yes 19 63.3Mizukietal.(1984) 1823years 40 Yes 12 (day1), 18(day 2),22 (day3)30(day1), 45(day 2),55 (day3)Nakagawa(1988) 24 Yes 16 67%NakamuraandMukasa(1992) 2023years 24 Yes 12 50Takahashietal.(1997) 1825years 34 1 3%Yamaguchietal.(1981) 1825years 320 Yes 124 39%Yamaguchietal.(1985) 1828years 513 Yes 218 43%Yamaguchietal.(1990a,b) 1828years 677 Yes 293 43%3059years 85 Low 8 9%6079years 73 Low 6 8%D.J.Mitchelletal. / ProgressinNeurobiology86(2008)156185 163signal processing techniques are used, the reported incidence of musignicantly increases (Kuhlman, 1978; Pfurtscheller and Aranibar,1978;Tiihonenetal.,1989)anditisnowsuggestedthatmuisanormal rhythm in almost all, if not all subjects. Thus, it is probablyonly the expression of that rhythm in visually inspected scalp EEGthat is maximal during adolescence. The same may prove true forFM-theta. In studies where a frontal-midline theta component hasbeen identied using signal processing techniques and paradigmsknown to evoke FM-theta rhythms, the prevalence is high (Gevinset al., 1997; Smith et al., 1999). FM-theta was more prevalent in the1315 years age range in patients with various diagnoses includingepilepsy, braindamage, behavioural disorders, autonomic disorders,headache, and other problems (Palmer et al., 1976).There is reasonable evidence that the occurrence of FM-theta isdifferent for different age groups. However, FM-theta has a numberof other features that maybeworthexaminingat different agelevels.It has been found that, during the performance of a passive auditoryandoddballtask, adultshavestrongerphase-lockinginthethetarange recorded at Fz, Cz, and Pz. However, the phase locking for theearly theta component was not as pronounced at Fz and Pz.Furthermore, adults tendedto showa greater increase intheamplitudeofthetafrompre-topost-stimuluspresentationthanchildren. Finally, children had signicantly higher theta amplitudesthanadults(YordanovaandKolev, 1997b).Greaterphase-lockingandamplitudeincreases havealsobeennotedinmiddleagedcompared to young adults (Yordanova et al., 1998). Further studiesstill needtoexaminewhat functional or performancevariablescoincide with these changes in FM-theta with age.5. Behavioural modulation5.1. EventsthatmodulateFM-thetaTheta rhythmicity can be present as a tonic background presentoverconsiderableareasof thescalpforconsiderableperiodsoftime. However, it can also occur in a phasic fashion, showing a clearincreaseinpowerlinkedtotheoccurrenceof eitherinternal orexternal stimuli. Preliminary attempts to assign functions to theta(bothintheolder rat hippocampal literatureandinthemorerecent human frontal literature) have focussed on the correlationbetween the phasic elicitation of theta (or changes in its power orfrequency) and specic stimuli or behaviours. However, ininterpreting this literature, it should always be remembered thatcorrelationdoesnotprovecause. Inthecaseofrathippocampaltheta, bothitselicitationandhighlysignicantcorrelationswithpower andfrequencyoccur inrelationtobehaviours (suchasrunningdownastraightalley)thatareunaffectedevenbytotalremoval of the hippocampus. It follows that the occurrence of thetawill often be a sign of some function being executed in some part ofthebrainbutitsoccurrence ataparticularsitemayindicatenomorethanthereceipt bythat siteof anefferencecopyof theactivity elsewhere. It may be important for function in thestructure generating it only on that subset of occasions when thatstructureproduces functional output.Arellano and Schwab (1950) were the rst to report thepresence of theta recorded from frontal electrodes in a young manduring the performance of arithmetic problems. Since then anumberof behavioursandcognitiveprocessessuchasworkingmemory, spatial navigation, episodic memory, andmeditationhave been shownto elicit FM-theta. Typically, some formofdifcult sequential mathematical exercisesuchas theUchida-Kraepelin exercise of repeated addition isemployed to elicit FM-theta. Table 2 lists the activationprocedures used to elicit FM-thetain those studies. The common feature traditionally cited for thesetasks is non-specic, focused or sustained attention or concentra-tion,orresistancetointrusion.However,anumberofothertaskcomponents such as encoding, retrieval, performance, and practicehave been shown to inuence FM-theta activity. Ishihara and Yoshi(1972)claimedthatthequantityofFM-thetawasrelatedtotheamountandspeedofmentalworkandthatittendedtoappearTable2Resultsfrom papersdemonstrating modulationoffrontalmidline thetarhythm duringthe performanceofabehaviouraltaskInvestigators FMTR FMTFC Task Field MaxHz Amp mv DurationArellanoandSchwab(1950) Notspecied Abstractivementalwork(arithmeticproblems)Anteriortovertex,frontocorticalAftanasandGolocheikine(2001)(res) 3.777.54 Meditation(internalisedattention)/blissfulstateFrontal AFz, FzAsadaet al.(1999)(res) 6.4 50 3 s Serialaddition,serialsubtraction,recollectionofKanjicharacters,attentiontoslow breathingFrontocentral(vm, EEG,MEG)Fz,FpzBastiaansenet al.(2002b)(res) 3.97.9 Hzdec(IBP) Spatialdelayedresponsetask Frontal Bastiaansenet al.(2002a) (res) Theta-IAF (IBP) Sentenceprocessing Frontocentraltemporo-parietalBanquet(1973)(res) 57 Upto 100 12 s 47 Hz Transcendentalmeditation Frontal F3Brazierand Casby(1952)(res) 6 50 >1 s Mentalcalculation ?vertex ?vertexBrookingsetal.(1996)(res) 4.37.8 Hz Airtrafc simulator Frontal FzBurgessandGruzelier(1997)(res) Wordrecognitionmemory task Frontocentral Caplanetal.(2003)(res) 48 Hz Virtualvisual-spatialnavigation Frontal? VariableChungetal.(2002)(res) 57 Hz Visuo-spatialN-back task Frontal AFzCohenetal.(2007)(res) 48 Hz ProbabilisticreinforcementlearningtaskFrontal FzdeArau joetal.(2002)(res) 3.7 Hz Virtualvisual-spatialnavigationFrontalto posteriorpropogation?Deiberetal.(2007)(res) 48 Hz VerbalN-backtask Frontal FzDietletal.(1999)(res) 47.5 Hz Repetitiveelectricstimulus Frontal FzDoppelmayret al.(1998)(res) Theta-IAF? EpisodicrecognitionmemorytaskFrontal Ekstromet al.(2005)(res) 48 Hz Virtualvisual-spatialnavigation Frontal Ferna ndez etal.(2000)(res/ill) 5.46/6.24 Hz Sternbergverbalworkingmemory Frontal Gevinset al.(1979a)(res) 47 Hz Kohsblock design FrontalD.J.Mitchelletal. / ProgressinNeurobiology86(2008)156185 164Table 2(Continued )Investigators FMTR FMTFC Task Field MaxHz Amp mv DurationGevinsetal.(1979b)(res) 47 Hz Kohsblockdesign, mentalpaperfoldingFrontal Gevinsetal.(1997)(res) 6(57) high 0.52 s 57 Hz Visuo-spatialandverbalN-backtaskFrontocentral(vm) AfzGevinsetal.(1998)(res) 47 Hz Visuo-spatialandverbalN-backtask?Frontocentral FzGevinsandSmith(1999)(res) 57 Hz Visuo-spatialN-backtask Frontal FzGevinsandSmith(2000)(res) 57 Hz Visuo-spatialN-backtask Frontal FzGevinsetal.(2002)(res) 57 Hz Visuo-spatialN-backtask Frontal FzGrunwaldetal.(1999)(res) 48 Hz Hapticexploration FrontocentralGrunwaldetal.(2001b)(res) 48 Hz Hapticexploration Frontocentral FzHarmony etal.(1999)(res) 3.9and5.46 Hz Mentalcalculation Frontal Harmony etal.(2001)(res) 7.8 Hz Figurecategorization Prefrontal, anteriorcingulateAFzHashimotoet al.(1988)(ill) 7 75 15 s Uchida-Kraepelin ?Frontocentral FzIlan andGevins(2001) (res)(ill) 57 Hz Visuo-spatialN-backtask Frontal FzInanagaet al.(1983)(ill) 5 50 13 s Uchida-Kraepelin ?Frontocentral FzInouyeetal.(1994a)(res) 6.5 100200 >5 s Uchida-Kraepelin Frontocentral FzInouyeetal.(1994b) (res) 6.5 100200 >5 s Uchida-Kraepelin Frontocentral FzIraminaetal.(1996)(res) 15 s 67 Hz Mentalcalculation Frontocentral(EEG/MEG)IshiharaandYoshi (1972)(ill) 6.5 >100 >7 s Uchida-Kraepelin Frontocentral FzIshiietal.(1999)(ill) 6 50 >10 s Serialsevens,serialaddition,powersof3Frontocentral(vm, MEG)FzJensenandTesche (2002)(res) 78.5 Hz Sternbergmemory task Frontocentral(MEG) Kahanaetal.(1999)(ill) 48 100200 12 s Virtual visual-spatialnavigationDiffuse? (iEEG) Katayamaet al.(1992)(res) 57.5 50 2 s Music,playing andimaginingFrontocentral FzKlimesch etal.(1994)(res) 3.97.4 HzERD/ERS Episodicmemorytask Frontal Klimesch etal.(1996b)(res) 4.696.69 Hz,6.698.69 HzERD/ERSImplicitepisodicmemory task Fronto-central Klimesch etal.(1997)(res) 4.36.3 HzERD/ERS Episodicmemorytask Frontal Klimesch etal.(2001)(res) Theta-IAF?ERD/ERS Episodicmemorytask Frontal FzKrauseet al.(2000) (res) 46 Hz,68 HzERD/ERSVerbal(visual)N-back task Anterior Krauseet al.(2001) (res) 46 Hz,68 HzERD/ERSModied auditorySternbergtaskAnterior Krauseet al.(2002) (res) 46 Hz,68 HzERD/ERSModied auditorySternbergtaskFronto-central Krauseet al.(2006) (res) 46 HzERD/ERS Visual > auditorystimuliinalexicaldecisionmakingtaskFrontal (diffuse) Kubotaet al.(2001) (res)(ill) 67 11.5 s 48 Hz Zenmeditation Frontal FzLaukkaetal.(1995)(res) 47 Hz Simulatedcar drivingtask Frontal FzLaukkaetal.(1997)(res) 47 Hz Simulatedcar drivingtask Frontal FzLehmann etal.(1993)(res) 17.5 Hz Abstract thinking Frontal Martin(1998)(res) 47 Hz Olfactorystimulation Frontal FzMacKay etal.(2001)(abstract) 47 Hz Visual-spatialnavigation/Sternbergmemory taskFrontal McEvoyetal.(2000) (res) 47 Hz Visuo-spatialN-backtask Frontal FzMcEvoyetal.(2001)(res) 47 Hz Visuo-spatialN-backtask Frontal Missonnieretal.(2006)(res) 47.5 Hz Verbal(visual)N-back anddetectiontaskFronto-central Mizukietal.(1980)(ill) (res) 6 75 17 s(20.5110 s) Uchida-Kraepelin ?Frontocentral FzMizukietal.(1983)a(ill) 6 75 14 s Uchida-Kraepelin ?Frontocentral FzMizukietal.(1984)(res) ? 310 s Uchida-Kraepelin Frontocentral FzMizukietal.(1986)(ill) 6 75 14 s Uchida-Kraepelin Frontocentral FzMizukietal.(1989)(res) 4.57.5 Hz Uchida-Kraepelin Frontocentral FzMizukietal.(1994)(ill) 6.57 75 14 s Uchida-Kraepelin ?Frontocentral FzMizukietal.(1996)(ill) 5.07.5 75 >1 s Uchida-Kraepelin ?Frontocentral FzMizukietal.(1997)(ill) 5.07.5 75 >1 s Uchida-Kraepelin ?Frontocentral FzMizutaniet al.(1985a)(abs) 67 Hz Uchida-Kraepelin ?Frontal ?FzMizutaniet al.(1985b)(abs) 67 Hz Uchida-Kraepelin ?Frontal ?FzMo lleetal.(2002) (res) 48 Hz Intentionallearningof words Fronto-temporal Moore etal.(2005)(res)(ill) 46 Hz Continuousgo/no-gotask Frontocentral ?FzNakamuraandMukasa(1990)(ill)67 75 23 s Uchida-Kraepelin ?Frontocentral FzNakamuraandMukasa(1992)(ill)67 75 23 s Uchida-Kraepelin ?Frontocentral FzNishikorietal.(1985)(abs) ? Mentalcalculation ?Frontal ?FzOntonetal.(2005) (res) 57 Hz Verbal(visual)Sternbergmemory taskFrontal FzPellouchoudetal.(1999)(res) 6.5 Hz Watching andplayingvideogameFrontocentral(vm) FzD.J.Mitchelletal. / ProgressinNeurobiology86(2008)156185 165clearlyifthetaskwasperformedsmoothly,butitdisappearedifthe calculation rate decreased. A similar nding was reported forthe MEG frontal-mental theta (Sasaki et al., 1996a, 1994a).Mizutani et al. (1985b) reported that task performance improvedduringperiodsofFM-theta, whencomparedtoperiodswhennoFM-thetawaspresent. FM-thetahasalsobeennotedtoincreasewithpractice during videogame playing(Smithet al., 1999).Mizutani et al. (1985a) reported an increase in performance speedof a mathematical task during periods when FM-theta was present.In these tasks FM-theta is usually elicited in discrete bursts of fairlyhomogeneousfrequency thatlastupto severalseconds,possiblywith an underlying period of around 4050 s (Mizuki et al., 1980).In our reviewof this literature, we will rst explore thesimplicationthat there couldbe anisolatedunitary frontal-midline theta responsible for a varietyof distinct behaviouralprocesses. Across behavioural studies the specic temporal,topographic, andfrequencycomponents of thetaactivityvary.Thereisalsoevidencethatthetaactivityatthefontalmidlineiscoupled with activity at other brain areas, not only at thetafrequencies(Sarntheinet al., 1998) but alsoinotherfrequencybands suchas gamma (Schack et al., 2002). The EEGstudiesdiscussedbelowsuggest onlythat FM-thetahas somekindofmnemonic or cognitive function but they neither make this certain,nor clearly identify a specic function that can be attributed to it;norshowwhetherthereisjustonefunctionaltypeofFM-theta.Experiments thus need to be carefully designed to tease outfunctionsof theta. Thefollowingsectionexaminessomeof thetasksthathavebeenrelatedtoFM-thetaactivity.Alatersectionwill discuss therelationshipbetweenFM-thetaandactivityatother brainsites.5.2. WorkingmemorystudiesWorking memory is the ability to temporarily hold andmanipulate a limited amount of information in immediateawarenessandisoftennecessaryinordertoproduceacorrectresponse(Baddeley, 1986;BaddeleyandHitch, 1974). Baddeleyhas proposedthat workingmemoryis comprisedof a centralexecutivecomponent, avisual-spatial scratchpad, andaphono-logical loop. The Sternberg task (Sternberg, 1966) and the N-backtask are frequently used to study working memory. The Sternbergtaskinvolvesthepresentationofastimuluslistcapableofbeingheldinworkingmemory, followedbyadelayperiodbeforethepresentationofaprobestimulus. Theparticipanthastoindicatewhether or not the probe was presented in the stimulus list. The N-back task involves the continuous presentation of items, wherebytheparticipanthastoindicatewhethereachitemisthesameordifferent from an item presented previously. The following sectionoutlines thedifferent tasks andparameters showntoelicit orproducechangesinFM-theta.A clear relationship exists between working memory tasks andfrontal-midline theta (48 Hz) spectral activity. Both the Sternbergtask(JensenandTesche,2002;MacKayetal.,2001;Ontonetal.,2005) and the N-back task (Deiber et al., 2007; Gevins and Smith,Table2(Continued )Investigators FMTR FMTFC Task Field MaxHz Amp mv DurationPetersonandThaut(2002) (res) 48 Hz Visualcontinuous matchtosampleFrontal F3/F4Raghavachariet al.(2001)(res) 49 Hz Sternbergworkingmemory task ?Frontal(iEEG) Sarntheinet al.(1998)(res) 47 Hz Workingmemory task ?FrontocentralSasakietal.(1994b)(res) 57 Hz Successivesubtraction,musicalimaginationFrontocentral(MEG) Sasakietal.(1996a)(res) 57 Hz Successivesubtraction,musicalimaginationFrontocentral(MEG) Sasakietal.(1996b)(res) 57 Hz Mentalcalculation/abstractthinkingFrontal(MEG) Sasakietal.(1996c)(ill) 6.3 75 >4 s Powersof 3 Frontocentral(EEG,MEG)FzSausengetal.(2004)(res) 47 Hz Objectvisualworking memory Fronto-parietal Sausengetal.(2006)(res) 47 Hz Taskswitching workingmemoryFronto-posteriorSausengetal.(2007)(res) 47 Hz ComplexsequentialngermovementsFrontal 37 Hz Verbal/non-verbalSternbergtask Frontocentral FzSederbergetal.(2003)(res) 48 Hz Delayedrecalltask Frontoandright-temporal(iEEG)Shinomiyaet al.(1994)(res) 6.63 54 1.97 s Spontaneouslyoccurring Frontocentral FzSmithetal.(1999)(res) 6.5 Hz Visuo-spatialworkingmemoryparadigmand videogameFrontocentral(vm) FzSmithetal.(2001)(res) 67 Hz Flightsimulator Anterior/frontal FzStametal.(2002)(res) 26 Hz PictorialworkingmemoryretentionFrontoandposterior-temporal,parietal,occipitalSuetsugietal.(1998)(res) 5.07.5 75 >1 s Uchida-Kraepelin ?Frontocentral FzSuetsugietal.(2000)(res) 5.07.5 75 >1 s Uchida-Kraepelin ?Frontocentral FzTakahashietal.(1997)(ill) 6 50 12.5 s Uchida-Kraepelin Frontocentral FzWilsonetal.(1999)(res) 4.37.8 Hz Workingmemory Frontal Yamaguchietal.(1990a)(abs) 67 Hz Mentalwork Frontal-midline FMTR implies FM-theta rhythm as identied from EEG records such as gures and descriptions, FMTFC implies FM-theta frequency components as identied using signalprocessing,res implies data obtained from results section. ill implies data measured from illustrative example, abs implies abstract. Theta-IAF implies theta adjustedaccording to individual peak alpha frequency. wm implies working memory. ? implies uncertainty because of insufcient data. (vm) implies that eld is estimated from avoltage map. (EEG) implies that eld is estimated from an illustration of raw EEG, iEEG implies intracranial EEG, MEG implies magnetoencephalography ERD implies eventrelated desynchronization, ERS implies event related synchronization, IBP implies induced band power. Max indicates a clear focal increase in FM-theta, whereas left blankindicatesnot enoughdata.aSamepatientsasInanagaet al.(1983)paper.D.J.Mitchelletal. / ProgressinNeurobiology86(2008)156185 1661999; Gevins et al., 1998, 1997, 2002; Krause et al., 2000;Missonnieretal., 2006;Smithetal., 1999)havebeenshowntoelicit FM-theta; as has relational list learning (Caplan and Glaholt,2007) andword-colour association(SummereldandMangels,2005). In addition, other tasks with a proposed working memorycomponent suchasmental rotation, Kohsblockdesign, serialaddition, substitutionof letterswithsubsequent wordrecogni-tion (Gevins et al., 1979a,b), Uchida-Kraepelin task (Ishihara andYoshi, 1972), ight simulation tasks (Smith et al., 2001), air-trafccontrol tasks (Brookings et al., 1996), videogame playing (Pellou-choud et al., 1999; Smith et al., 1999), the playing of complicatedpiecesofmusic,ortheimaginingofthatmusic(Katayamaetal.,1992), andcomplexandnovel motor sequencetasks (Sausengetal., 2007)alsoproducethetaatfrontallocations. TheworkingmemorytasksshowntoincreaseFM-thetainvolveanumberofdifferent sensory modalities including motor (Sauseng et al., 2007),verbal(Gevinsetal.,1997;Krauseetal.,2000),auditory(Krauseetal., 2001), andvisual-spatial(GevinsandSmith, 1999;Gevinset al., 1997) information. Gevins et al. (1997) found that there werenodifferencesintheamountof FM-thetaproducedbyeitheraverbal orvisual-spatial task.Perhaps the most often reported relationship shows an increasein the amount of 47 Hz FM-theta with increasing workingmemoryload, task difculty,or mentaleffort (Gevinsand Smith,1999, 2000; Gevins et al., 1998, 1997; Grunwald et al., 1999; Jensenand Tesche, 2002; Laukka et al., 1995; McEvoy et al., 2000; Sausenget al., 2007; Smithet al., 2001; Yamaguchi et al., 1990b). Forexample, JensenandTesche(2002)usinga122channel, wholehead, MEG array during a Sternberg paradigmfound that FM-thetaintherangeof78.5 Hzincreasedinpowerasafunctionofthememory set size and persisted during retention and comparison oftheprobetothememoryset. Consistentwiththis, Ontonetal.(2005) demonstrated an increase in 57 Hz theta activity at Fz withanincreaseinworkingmemoryloadduringtheSternbergtask.Interestingly, the increase in FM-theta with working memory loadwas the result of a small fraction of trials showing a large increaseinactivityratherthanamodestincreaseonmosttrials.Thishasimportant implications both for methodology and interpretation.Recent attempts have also been made to separate outcomponents of workingmemoryintheEEG. Missonnier et al.(2006) investigated 47.5 Hz ERS during a verbal N-back 1 task, anN-back 2 task, a detection task, and a passive xation task that eachinvolved the visual presentationof verbal stimuli. There was a clearincreasein47.5 HzERSfollowingstimuluspresentationduringeachofthetasksthathadashorterlatencytopeakatposteriorcompared to anterior electrode sites. At frontal electrode sites theamplitude of the peak was greatest during the detection, followedbytheN-back1, N-back2andthenpassivetask. Thegreatestamplitude occurredduring the attentiontask whichwas alsosignicantly greater than the 47.5 Hz peak that occurred duringtheN-back2andpassiveconditionsatbothcentral andfrontalelectrodes sites. Thus, theta observed during such tasks may reectthe allocation of attentional resources related to working memory.Missonnier et al. (2006) suggested that the lower level of spectralpowerobservedduringtheN-back2conditionmayreect theallocationofresourcestootheraspectsoftheworkingmemorytask. Similarly, Deiber et al. (2007) found that 48 Hz energy wasgreater followingstimulus presentationintheattentional andN-backtaskscomparedtothepassivetaskat frontal electrodesites. Deiber et al. (2007) alsoreportedontheta componentsrecorded at Fz to examine components of working memory using asimilarparadigm. Greater48 Hzenergywasnotedduringthedelayperiodonno-responsetrials duringtheverbal N-back2comparedtoN-back1anddetectiontasksatfrontal electrodes.Transient energy, occurring earlier in the delay period, wasreducedinthepassivetaskbutwassimilarfortheN-backanddetection tasks. These results suggest that earlier transientcomponents of theta at frontal sites may be important forattention-relatedfunctionswhereaslatersustainedcomponentsare important for working memory. This is consistent with Sausenget al. (2007) who suggested that frontal-midline theta per se mayreect theactivityof anattentional systemwhilefrontal thetacoherencewiththetaatotherbrainregionsmaybeinvolvedindealingwithworkingmemorydemands suchas theexecutivecontrol andintegration ofinformation.FM-theta has also been related to additional performanceparameters. Gevins et al. (1997) found that reaction speeddecreased, accuracy increased and the FM-theta responseincreasedneartheendof practice. Smithet al. (1999) demon-strated that the amount of FM-theta increased over time and oversessionsinbothanN-backexperimentandaspatial navigationvideogame.Ferna ndezetal.(2000)lookedatthetainrelationtocorrect andincorrect answersduringaverbal (Sternbergpara-digm)workingmemorytask, spatialworkingmemorytask, andduring calculation of mathematical sums. Prior to correctresponsesmadeontheverbalworkingmemorytasktheynoteda statistically signicant increasein 5.46 Hz EEG powerrecordedfromthescalpregionsoverlyingtheleftdorsolateral prefrontalcortex. In contrast, increases in 6.24 Hz EEG power were recordedat sites over the anterior cingulate and right frontal cortex prior toincorrect answers on the mathematical task. Harmony et al. (2001)observed an increase in the power of theta (7.8 Hz) recorded overprefrontal, anterior cingulate, andanterior temporal regions inchildrenwhileperformingagurecategorizationtask. Harmonyet al. (1999) found that the 5.46 Hz frequency was more prominentin the right pre-frontal regions during mental calculation than incontrol conditions. During a verbal N-back task Krause et al. (2000)found greater 46 Hz ERS at anterior electrodes when targetstimuli had been presented on previous trials. Caplan and Glaholt(2007) found that, with relational list learning in which presenta-tion of one word from a previously presented list cued recall of aprecedingor followingword, FM-thetawas linkedtomemoryperformanceandwassomoreinrelationtoindividualvariationacrossparticipants thanacrosstrialswithin participants.FM-theta has also been shown to change dependent ondifferent components of thetask. As notedabove, JensenandTesche(2002) reportedtheappearanceof FM-thetaduringallfacets of the Sternberg task. However, Peterson and Thaut (2002)noted greater 48 Hz spectral power at F3 and F4 during 5 and 10 sdelay periods compared to 2 s delay periods in a continuous non-verbal auditory working memory task relative to a referencememory task. Furthermore, the context in which tasks arepresented during an experiment may be enough to producechangesinFM-theta. Wilsonet al. (1999) foundthat FM-thetaincreased during the retention period with task difculty during aworkingmemoryexperiment, butonlyinonecondition. Duringcondition1, participantshadtoremember1, 3, 5, 7, or8items,retain themfor 4 s and then recall. For this condition the easiest listlength of 1 itemwas presented on 60% of occasions, and each of theotheritemsmadeup40%of thetrials. Subsequent testingwasconducted where memory sets for each difculty level wererandomlypresented(condition2)orpresentedinblocksof 40-trials (condition 3) during counterbalanced experiments. Nochanges in theta were noted at frontal electrodes for the differentdifcultylevelsduringthissubsequent testing. ThetaERSoveranterior, central, and posterior electrode sites during an auditoryversionoftheSternbergtaskwasshowntobeconsistentattwodifferenttimepointsspacedapproximately9daysapart(Krauseet al., 2001). Relatively high Pearson correlation coefcientsbetweentestsessionswerenotedinthethetafrequencyrangeD.J.Mitchelletal. / ProgressinNeurobiology86(2008)156185 167duringboth encodingandretrievalcomparedtootherfrequencybands. McEvoy et al. (2000) alsodemonstratedthat FM-thetashowed reliability during an N-back task both within and betweensessions. Sausengetal. (2004) foundthat47 Hzthetaactivityrecordedfrompre-frontal sitesduringatwopartencodingandretrievalworking memory task.It has also been suggested that the phase of theta may be resetfollowing stimulus presentation in working memory tasks.Researchconductedby Luuandco-workers (LuuandTucker,2001; Luu et al., 2004) has implicated frontal-midline thetacomponentsasbeingresponsiblefortheerrorrelatednegativitywhich is observed during the monitoring of errors. A similar phasereset of hippocampal thetatoasimplestimulusoccurringinaworking memory butapparently not inareference memory taskhas been reported in rats (Givens, 1996). However, phase reset canoccur (but perhapstoagenerallylesserextent) withreferencememory (Williams and Givens, 2003) and may be a generalmechanismfor controllingpreferential processingof incomingstimuli(Buzsa ki,2006,Cycle 10).Themajorityofthestudieshaveshownanincreaseinthetapower or activity during working memory tasks. However,decreasesorabsencesof changeintheactivityof thetaduringtask performance could also be useful for unravelling the functionof this rhythm. Bastiaansen et al. (2002b) looked at the effects ofperformingadelayed-responsespatialworkingmemorytaskontheta rhythm. During these trials a circle stimulus was presented toperipheral vision in relation to a centre square xation point. Thiscircle disappeared after 150 ms during memory trials butremained on sensory trials. The xation point was removed either1150 or 4150 ms following trial onset. When the xation point wasremoved the participant had to indicate where on the screen thestimulus was presented. What resulted was an increase in inducedtheta andlower-1 alpha (averaged over 3.97.9 Hz) over parietalregions during the display of the stimulus. During retention therewasadecreaseininducedbandpowerinthethetaandlower-1alphaoverfrontal electrodes. Furthermore, duringthememorytrials there was a greater reduction in the theta frequency powerrecorded over frontal electrodes. Here it was suggested that,depending on the type of function being carried out, certainhippocampo-cortical loops are activated. In this instance, theta inthe hippocampo-frontal loop is decreasedin effect reducing noisethatmayinterferewithperformanceonthetaskthatinvolvesahippocampo-parietal loop. Similarly slowwaves over frontalregionswerenotedduringretrieval ofverbal informationwhilespatial information resulted in an increase in slow waves,particularlyinparietal regions(Heil et al., 1996, 1997). Thisissimilar tothendingbyPesonenet al. (2007) where 46 Hzactivityincreasedduringavisual N-backtaskparticularlyoverparietal regions. Finally, Raghavachari et al. (2006) found that thegating of theta power, or amplitude enhancement was essentiallyabsentatfrontalintracranialcorticalsites(seebelowfordetailsregarding intracranial recordings) for the durationof the trialperiod during a verbal Sternberg task. However, it was difcult togauge whether these electrode locations were near frontal-midlinesites.5.3. SpatialnavigationtasksSeveral studies have looked at theta in relation to virtual spatialnavigation tasks. MacKay et al. (2001) examined theta during botha Sternberg verbal working memory task and a virtual maze task.They found that theta tended to be distributed over frontal regionsfor bothtasks, while the working memory task also hadlefttemporal thetaactivityandthevirtual mazetaskdemonstratedconcurrent right temporal activity. A number of spatial navigationtasks have involved the use of intracranial EEG (iEEG) in patientswithepilepsy.Kahana et al. (1999), using iEEGin three patients, foundincreases in theta activity during long versus short mazes, notablyover temporal sites. Caplan et al. (2001) examined iEEG inparticipantswithepilepsywhilelearningamaze. Thetarhythmwas the predominant activity that was recorded during thelearningofthemaze.Furthermore,asthemazelengthincreasedthe occurrence of theta rhythm also increased. Theta rhythm waswidely distributed across the cortex. Caplan et al. (2003) found thattheta activity during exploration and goal seeking behaviour in avirtual navigation task had different cortical distributions. Inparticular widespread theta was noted over motor, temporal, andfrontal cortices in Fig. 4 of their data. Fig. 7 of their data revealedgreater theta power during the goal seeking task versus theexploration task at a number of sites including the frontal cortex.Ekstromet al. (2005) foundanincreaseinthetapower duringvirtualnavigationatcorticalandhippocampalelectrodesites.Atcortical sitesthetarecordedfromfrontal andtemporal regionsduring virtual navigation was greater when searching forrandomlydistributedtargetscomparedtosearchingfortargetsthat had xed locations. Furthermore, the amount of cortical (non-specic) theta recorded correlated with the amount of thetarecorded fromthe hippocampus during segments of navigation. deArau jo et al. (2002) using MEG, found a peak in the 4 Hz range thatincreasedduringnavigation. Usingasingledipolemodel theylocalized this source to near the superior temporal gyrus and thedeepertemporalstructures(p. 72). Theauthorsalsofoundthattheta activity propagated fromfrontal to posterior regionsduringthe task.There are a number of advantages and disadvantages to usingiEEG(Caplanetal., 2001;OKeefeandBurgess, 1999). Themainadvantages are that it is possible to determine with more precisionthe source of the underlying rhythmat the same time asminimizingmuscleandeye-blinkartefacts. However, theiEEGrecordings are restrictedtopeople inwhomthere is possibleabnormal EEGand abnormal brain function. Thus, there is asampling bias with greater numbers of electrodes being placed intemporal lobes (Caplanet al., 2001). Furthermore, patients areoften taking medications and even if they are not, and recording isfromareas with no explicit pathology, the tissue recorded fromwillpotentiallyhavebeensubjectedtoabnormal inputfrompatho-logical areasover considerableperiodsof time. Finally, clinicalapplications of iEEG only extend to certain population groups andpatient numbersareoftensmall.5.4. EpisodicmemorytasksAnumberofstudieshavereportedthepresenceofthetaERSactivity at frontal electrode sites during the performance ofepisodic memory tasks. A prominent nding of episodic memorystudiesisthatthetaincreasesduringencoding(Klimesch, 1999;Klimeschetal., 1997). Furthermore, thetarecordedfromfrontaland other cortical regions during encoding is related to successfulretrieval (Klimeschetal., 1994;Osipovaetal., 2006; Sederbergetal.,2003).Similarly,Mo lleetal.(2002)reportedthateffectivelearning or encoding of pairs of words was related to asynchronizationof thetaandadesynchronizationof alphaoverthe frontal and temporal regions of the scalp. Klimesch et al. (1994)conducted a study where participants had to later indicate (whichparticipants were not informed about) whether a congruentconcept-feature(e.g. eagle-claw)pairhadbeenpresentedinanearlier semantic task. In the rst 375 ms of the pre-stimulus periodthe power of theta was greatest at frontal electrodes. Following thepresentation of a stimulus in the episodic memory task where pairsD.J.Mitchelletal. / ProgressinNeurobiology86(2008)156185 168had been presented earlier, there was an increase in relative thetapower. This tendedto decrease at about 375 ms followedbyanotherincreaseat750 msthatwasmostpronouncedatfrontalelectrodes.Thereisalsoanincreaseinthetafollowingrecognitionandretrieval. Klimesch et al. (1996a) conducted a study that involvedthepresentationof 96words that participants hadtoinitiallycategorizeaslivingornotliving.Withoutpriorwarning,laterintheexperimentparticipantswererequiredtorecall asmanyofthesewordsaspossible.Whatwasfoundwasanincreaseintheabsolute power andevent relatedpower (over time) of thetaduring encoding for words that were later recalled correctly.Furthermore, the power of theta was signicantly higher in the lefthemisphereandatcentral andfrontal electrodesites. Klimeschetal. (1997)specicallylookedatthetasynchronizationduringencodingandretrievalinvolvedinanepisodicmemorytask.Forcorrect trials, both encoding and retrieval components of the taskwereassociatedwithanincreaseinthepowerofthetathatwasmost pronouncedover frontal regions. Alater study(Klimeschet al., 2001) demonstrated that retrieval compared to encoding wasassociatedwithasignicantlygreatersynchronizationof theta.Thiswas mostpronounced atfrontal regions.Thetasynchronizationover frontal-midline regions has alsobeenshown to be twice as large following presentation (recognition) ofwords that had previously been presented (Burgess and Gruzelier,1997). Similarly, Doppelmayr et al. (1998) investigated ERDandERSduring an episodic recognition memory task that involved having tocorrectlyidentifypreviouslypresentedtarget stimuli fromdis-tractor words. There was greater theta synchronization at allelectrodesitesintheright hemisphereduringretrieval ingoodperformers than bad performers. For the left hemisphere, there wasa greater degree of theta synchronizationinfrontal electrodescompared to central, temporal, parietal, and occipital electrodes.5.5. InternalisedattentionandmeditationSeveral studies have demonstratedthat FM-theta increasesduringperiodsofsustainedinternalisedattentionormeditation(Aftanas andGolocheikine, 2001; Banquet, 1973). Aftanas andGolocheikine(2001) notedanincreaseintheta(46 Hz) andalpha 1 (68 Hz) power over frontal and midline regions in long-term meditators compared to controls and short-term meditators.There was also a synchronization of theta between prefrontal andposteriorassociationcortexwithafocalpointlocatedintheleftprefrontal cortex. Kubotaet al. (2001) alsoconductedastudywhereparticipants performedameditativetechnique(Su-sokotask) that involved focusing on the breath and counting thenumber of inhalations and exhalations. Performance of thismeditative technique produced an increase in the amount ofFM-theta (consistent with other meditation studies). Furthermore,the increase in FM-theta correlated with autonomic heart activityincludingatrendfor increasedinter-beat interval, increasesincardiac vagal and sympathetic indices. In contrast, cardiacsympathetic activitywas inverselyrelatedtoFM-theta. Asadaet al. (1999) alsoreport FM-thetawas consistentlyelicitedbyrecollectionof Kanji charactersandattentiontobreathingveryslowly in their subjects who were recruitedbecause of theirreliable FM-theta. Furthermore, a blissful meditative state attainedduringinternalisedattentioncoincidedwithanincreaseinFM-thetapower.5.6. PracticalandothertasksthatmodulateFM-thetaA number of more ecological tasks have been shown to produceFM-theta. Brookingset al. (1996) observedthetarhythminairtrafccontrollerswhileperformingsimulatedairtrafccontroltasks. High difculty compared to lowdifculty resulted inincreased theta recorded from F8, C3, CZ, T4, P3, Pz and P4. Duringanoverloadconditiontherewas signicantlymorethetaoverfrontalregionsincludingFz. Similarlyinaightsimulationtask,FM-thetaincreasedasthetaskwasmademoredifcult(Smithet al.,2001).Laukka et al. (1995) investigated how FM-theta changes duringa driving task. EEG was measured at different stages during eachdriving trial. At two points in the trial the participant had to choosebetween different roads on which to travel. The correct road to godownatpoint2wasdependentontheroadchosenatpoint1.Participantshadtolearnandthenpracticethediff