Anda di halaman 1dari 18

ARTICLE IN PRESS

ARTICLE IN PRESS Clinical Neurophysiology xxx (2006) xxx–xxx Invited review www.elsevier.com/locate/clinph Clinical

Clinical Neurophysiology xxx (2006) xxx–xxx

Invited review

Clinical Neurophysiology xxx (2006) xxx–xxx Invited review www.elsevier.com/locate/clinph Clinical neurophysiology of

www.elsevier.com/locate/clinph

Clinical neurophysiology of language: The MEG approach

Riitta Salmelin *

Brain Research Unit, Low Temperature Laboratory, Helsinki University of Technology, Espoo, Finland

Accepted 28 July 2006

Abstract

Clinical evaluation of language function and basic neuroscience research into the neurophysiology of language are tied together. Whole- head MEG systems readily facilitate detailed spatiotemporal characterization of language processes. A fair amount of information is available about the cortical sequence of word perception and comprehension in the auditory and visual domain, which can be applied for clinical use. Language production remains, at present, somewhat less well charted. In clinical practice, the most obvious needs are noninvasive evaluation of the language-dominant hemisphere and mapping of areas involved in language performance to assist surgery. Multiple experimental designs and analysis approaches have been proposed for estimation of language lateralization. Some of them have been compared with the invasive Wada test and need to be tested further. Development of approaches for more comprehensive pre-sur- gical characterization of language cortex should build on basic neuroscience research, making use of parametric designs that allow func- tional mapping. Studies of the neural basis of developmental and acquired language disorders, such as dyslexia, stuttering, and aphasia can currently be regarded more as clinical or basic neuroscience research rather than as clinical routine. Such investigations may even- tually provide tools for development of individually targeted training procedures and their objective evaluation. 2006 International Federation of Clinical Neurophysiology. Published by Elsevier Ireland Ltd. All rights reserved.

Keywords: Speech perception; Reading; Speech production; Picture naming; Magnetoencephalography

1. Intr oduction

In the clini cal con text, the que stion of language repre- sentat ion in the human brain is largely focused on pre-s ur- gical map ping, in the form of noninvasi ve laterali zation of language functio n a n d c haracteriza tion of linguist ic pro - cesses that are repres ented in the imm ediate vicinity of an area to be resected . Anothe r topic of clinical inter est is tracki ng of neural e ects associated with rehab ilitation after acquire d language disorde rs, or wi th specific training program mes in the case of de velopm ental disorde rs. All these ap plications critically depend on, or sh ould de pend on, infor mation de rived from basic resear ch of the organi- zatio n o f language function in the human brain. The success ive an d large ly overlap ping stage s i n lan- guage process ing, from sensory analys is (visu al, auditor y,

* Tel.: +358 9 4512950; fax +358 9 4512969. E-mail address: riitta@neuro.hut.fi.

and tact ile) to lingui stic asses sment , memor y search, an d motor function , can only be satisfa ctori ly ch aracterize d and unde rstood using combined spatial and tempor al i n f o r m a t i o n . W h o l e - h e a d m a g n e t o e n c e p h a l o g r a p h y (MEG) lends its elf as an obviou s tool in this end eavour as it allows fast tracki ng of brain a ctivatio ns at milli second time resolution and reasonabl e spatial accu racy. This pa per discus ses the neu ral organiz ation of language function as it appears in MEG recordi ngs. We wi ll first out- line ne ural pr ocesses of sp eech percept ion, reading , and speech produ ction in healt hy subjects and then pro ceed to consider how this infor mation may be used in the clinical dom ain.

2. Speech percept ion

2.1. Sequence of acti vation

Speech percept ion is though t t o pr oceed as foll ows:

speech signal s enter the ear as sound waves from whi ch

1388-2457/$32.00 2006 International Federation o f Clinical Neurophysiology. Published by Elsevier Ireland Ltd. All rights reserved.

doi:10.1016/j.clinph.2006.07.316

ARTICLE IN PRESS

2

R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx

the br ain extra cts speech soun ds an d spee ch sou nd sequence s whi ch furt her activate the meani ng of the word ( Hickok an d Poeppe l, 2004 ). At the level of the brain, activity is detect ed in the superi or tempor al cortex bilateral ly ( Fig. 1 ). All sounds evoke a pro minent activati on at about 100 ms, referred to as the N100 response, or N100m in MEG ( Hari, 1990 ). Wh en listening to natural sp eech, words , a n d sen- tences , the N100m response is foll owed by a susta ined activati on that star ts at abo ut 200 ms after stimu lus

onset, reaches the maxi mum at ab out 400 ms, and lasts until 600–800 ms (e.g ., Bierm ann-Ru ben et al., 2005; Hel- enius et al., 2002b; M arinkovi c et al., 2003 ). This acti va- tion is often referred to as the N400 response, or N400m in MEG. From the MEG (or EEG) point of view, with the lim - ited spatia l resolu tion (1–2 cm), all this acti vity is g ener- ated in the same general cortical area. How ever, based on a numb er of experi ments , there are various pro cessing stage s reflect ed in this activatio n, at di erent times. Acoust ic–phonet ic feat ures of speech mo dulate acti vity in non- primary auditory cortex from 50–1 00 ms onwar ds, as reflect ed in the N100m response ( Kurik i and M urase, 1989; Obleser et al., 2004; Par viainen et al., 2005; Poep- pel et al., 1996 ). Lang uage-s pecific phonetic– phonol ogical analys is starts by 100–200 ms afte r stimulus ons et. In this time win dow, an MEG/E EG response associ ated with mnemoni c function s o f the auditor y associ ation cortex,

Na¨ a¨ ta¨ -

i.e., the mismat ch negativit y (MMN , i n EEG ; cf.

nen et al., 2005 ) o r mis match field (MMF , in MEG) indica tes acce ss to phonol ogical categor ies ( Phill ips et al., 2000; Vi hla et al., 2000 ), and dist inct proc essing of native vs. nonna tive phone tic con trasts ( Na¨ a¨ ta¨ nen et al., 1997 ). From about 200 ms onwar ds, the superi or tempor al activati on shows sensitiv ity to lexical- semantic manipul ation (Hele nius et al., 2002b ; K u jala et al., 2004; M arinkovi c et al., 2 003 ). How may one extract this

type of tim ing infor mation ? Let us consider exampl es of experi ments that have been used for ch aracterizi ng the stage s o f acoustic/ phonetic analys is (N100m tim e win- dow) and lexic al-semant ic pro cessing (N40 0m time wind ow).

2.2. Acousti c–phonetic analysis: exampl e

Par viainen and colle agues (Par viainen et al., 2005 )

asked whet her speech- specific analys is is reflect ed in the neural process ing before the MMN/ MMF time wind ow,

in the N100m respo nse. The stimuli were synthet ic vow-

els /a/ and /u/ and consonant- vowel syll ables /pa/ an d /

ka/, an d c o rrespondi ng non-speech sound s. The complex non-speech sou nds contai ned the dominant frequenci es of the speech so unds (three forman t frequenci es, F1, F2, and F3), an d the simple non-speech sounds one for- mant frequency (F2). For vowels, the frequency content remains essent ially the same throu ghout the soun d whereas for the conson ant-vowel co mbinations there is

a rapid frequency trans ition at the be ginning . I n this

study, synthet ic sounds wer e used (as opposed to na tural speech) in order to control for aco ustic propert ies as well as possible. The stre ngth of the N100m response in the left hemi- sphere was sensitiv e t o the stimulus content , with stron- gest activati on to sp eech sounds and weake st to simple non-speech sounds, sim ilarly for vowel s and consonan t- vowel syllables ( Fig. 2 a). No such depend ence on stimu- lus type was detected in the right he misphe re. Both the left and right superi or tempor al co rtex were obviousl y involved in pro cessing all so unds but the variation in the left he misphere resul ted in a leftw ard shift of he mi-

spheric ba lance for speech sound s. The amp litude behav- iour alone does not necessa rily indica te a specia l r o le for speech so unds as the signal stre ngth could refle ct an

e ect of systemat ically increa sing acoustic co mplexity of

the stimuli, not speech- specificity pe r se. Ho wever, the timing of activatio n provides addition al relev ant informa- tion (Fig. 2 b). For speech sounds, the build- up of the N100m respon se was signi ficantly faster in the left than right hemis phere wher eas for the non-spee ch sounds, simple or complex , n o such di erence in the ascendi ng slope was observed. Spe ech seems to be specia l b y 100 ms a fter stimulus onset. Thes e findin gs a r e in line with studi es that ha ve shown stronger N100m amp litude for vowels than piano notes

or tones (Go otjes et al., 1999 ), long er late ncies for v owels

( Go otjes et al., 1999 ), long er late ncies for v owels 0 200
( Go otjes et al., 1999 ), long er late ncies for v owels 0 200
0 200 400 600 800 Time (ms) Lexical-semantic analysis Phonological categories Acoustic/phonetic features
0
200
400
600
800
Time (ms)
Lexical-semantic analysis
Phonological categories
Acoustic/phonetic features

Fig. 1. Time course of speech perception in the superior temporal cortex, based on neurophysiological recordings.

ARTICLE IN PRESS

)mAn(htgnertsnoitavitcA)mAn(htgnertsnoitavitcA

Oxford University Press 2005

Oxford University Press 2006

R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx

3

a

b

/ Clinical Neurophysiology xxx (2006) xxx–xxx 3 a b 60 40 20 0 80 60 40

60

40

20

0

80

60

40

20

0

80

60

40

20

0

80

60

40

20

0

/a/ /pa/ /a/ /pa/ Speech Complex
/a/
/pa/
/a/
/pa/
Speech
Complex
Simple 60 80 100 120 60 80 100 120 60 80 100 120
Simple
60 80
100
120
60 80
100
120
60 80
100
120

Time (ms)

Fig. 2. Neural responses to speech and non-speech sounds. (a) Strength of the N100m response (mean + SEM) in the left and right superior temporal cortex to the /a/ and /pa/ stimuli and the corresponding complex and simple non-speech sounds. (b) Timing of the build-up phase of the N100m response, from onset to maximum, for the di erent stimulus types of the / a/ category. As the stimulus was given to the right ear the left-hemisphere response is systematically about 10 ms faster. Modified from Parviainen et al. (2005).

than tones (Eul itz et al., 1995; Tiitinen et al., 1 999 ), or left- ward shift of he mispheric balance for natural vowel s as compared with co mplex tones ( Vihla an d Salme lin, 2003 ). Apart from the study of Parviai nen and colle agues (Par - viain en et al., 200 5 ), the sti muli ha ve typicall y been sound s with stable frequenci es (i.e., vo wel type soun ds) ( Eulitz et al., 1995; Tiit inen et al., 1999; Vih la and Salme lin, 2003 ) o r trans ition sounds (i.e., conso nant-vo wel syllab le type of sounds) ( Shtyrov et al., 2 000 ) but not both . As nat- ural language is a mixt ure of these sound types, it may be impor tant to allow acousti c variation amon g the speech stimuli when evaluating corti cal an alysis of spe ech vs. non-speech sou nds.

2.3. Lexical -semant ic analys is: example

Here, we move from syllables to complet e words , and to extracti on of meani ng in reali stic speech percep tion. Bonte and colleagu es ( Bonte et al., 2006 ) ex plored emerg ence of lexical- semantic analys is in the analys is of spoken words . Their stimulus set was comp osed of a large number of nat- ural sp oken sentenc es from which both the initial words and the initial syllables were cu t a s separat e stimuli . The initial syllables are pot entially meani ngful but only if fol- lowed by furt her speech. The pe rception of syllab les was tested in two con texts. One sequence contain ed onl y sylla- bles an d, theref ore, expecta tion for meani ngful language was low. Anothe r sequ ence included all stimulus types, syl- lables, words , a n d sentenc es, played in a random orde r. Accor dingly, expecta tion for meanin gful language was high as any syllable could signal the be ginning of a sentenc e or word. This study thu s varie d the need for semant ic a nal- ysis while eq uating phone tic/phonol ogical aspect s b e tween the expe rimental con ditions. Fig. 3 displays the tim e course of activati on in the au di- tory cortex to syllables played in isolati on and to syllab les played in context of words and sentenc es. The patte rn was significan tly di erent from abo ut 200 ms onwar ds. For syl- lables in context , the su stained acti vation star ted earlier and was overal l strong er than for syllab les in isol ation. In the ascend ing slope, the acti vation evok ed by syll ables in context coinci ded wi th that ev oked by words and sentenc - es. In potentia lly meani ngful context auditor y c o rtex thus seems to treat syllab les as words and sentenc es, just in c ase. By about 300 ms, the system had determ ined that there was no infor matio n t o follo w the initial syll able and the signals started to diverg e. Ext raction of mean ing, or potential meani ng in syllables , thus seems to star t at abo ut 200 ms after stimulus ons et.

Sentences Words Syllables (context) Syllables (isolation) 50 25 0 200 400 600 800 noitavitcA )mAn(htgnerts
Sentences
Words
Syllables (context)
Syllables (isolation)
50
25
0
200
400
600
800
noitavitcA
)mAn(htgnerts

Time (ms)

Fig. 3. Cortical e ect of expectation for meaning. Time course of activation in the left superior temporal cortex for spoken sentences, sentence-initial words, and sentence-initial syllables. The syllables were presented either in context, as part of a sequence composed of all stimulus types, or in isolation, as a separate sequence. Modified from Bonte et al.

(2006).

ARTICLE IN PRESS

Society for Neuroscience 2002

4

R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx

An e cient setup for cha racterizin g n e ural co rrelates of lexical- semantic analys is is to use sentenc es that create a very high expecta tion for a certain final word, as in The pia- no was out of tune , and break that expecta tion in a number of ways ( Connol ly and Phill ips, 1994; Helen ius et al., 2002b ); this experimen tal design was initially intro duced in the dom ain of reading ( Kutas and Hil lyard, 1980 ). Fig. 4 illustr ates the resul ts from a study by Hele nius and colle agues (Hele nius et al., 2002b ) wher e, in ad dition to the expecte d end ings, the sentenc es cou ld end with three dierent types of une xpected words . The sentenc e-endi ng word could be total ly anomal ous, as in The pizza was too hot to sing , or i t could be une xpected but have a meani ng that was plausi ble in the sentenc e context , a s in When the power went out the house becam e quiet when most people woul d expect dark . Yet another type was a word that sound ed correct in the beginni ng but had the wrong mean - ing, as in The gambler had a stre ak of bad luggage , instead of luck . The N100m respon se was not a ected by the stim- ulus type. The responses to the di erent sentenc es started to di er at abou t 200 ms. There was a promi nen t N400m response to the semant ical ly wrong end ings, a weake r acti - vatio n for the une xpected but semant ically plausi ble fina l words , a n d the weake st respon se to the exp ected words, thus resul ting in a pattern of activati on syst ematical ly g rad- ed by semant ic plausi bility. Di er entia tion be tween word types was de tected in both hemisp heres. Over all, the une x- pected sentenc e-endi ng words evo ked strong er activa tion in the left than right hemisp here. These data (Hele nius et al., 2002b ) and other MEG studi es using sentenc es or semantical ly congruent vs. incon - gruent word pairs agree on involv ement of the sup erior tempor al cortex in lexi cal-semant ic analys is of spo ken words , starting at ab out 200 ms afte r stimu lus onset ( Kuj- ala et al., 2004; Marinkovi c e t al., 2003 ). Thi s i s a co nsis-

Expected Meaning Sound form Anomalous Activation strength (nAm)
Expected
Meaning
Sound form
Anomalous
Activation strength
(nAm)
Meaning Sound form Anomalous Activation strength (nAm) 10 10 0 200 400 600 0 200 400
10 10 0 200 400 600 0 200 400 600 Time (ms) Time (ms)
10
10
0
200
400
600
0
200
400
600
Time (ms)
Time (ms)

Fig. 4. Neural correlates of lexical-semantic processing in speech percep- tion. The black dot indicates the location of the auditory cortex and each white dot the centre of the active cortical patch during the N400m interval, in one individual. The curves display the mean time course of activation in the depicted source areas. The semantically wrong endings (totally anomalous, expected sound form at the onset) resulted in a prominent N400m response. The signal was smaller for the equally unexpected words that had a plausible meaning in the sentence context, and weakest for the expected words. Modified from Helenius e t al. (2002b).

tent finding regardl ess of wheth er an acti ve ne uronal populati on is mod elled as a focal Equivale nt Curr ent Dipole (ECD; Ha¨ ma¨ la¨ inen et al., 1993 ) whi ch repres ents the centre of an acti ve corti cal patch a n d direct ion an d magnitud e o f elect ric current therei n, or displ ayed as a dis- tributed pro bability map (e.g ., Dale et al., 20 00; Uu tela et al., 1999 ). Dis tributed source mod elling of MEG data (M arinkovic et al., 2003 ) suggests that neural activit y underlyi ng the N400m response may add itionally extend into (left) anterior tempor al an d front al areas. It is impor - tant to reali ze that focal ECDs and distribut ed probabil ity maps prod uce exactl y the same electromag ne tic fie ld out- side of the hea d s o they are both equall y correct models of the unde rlying neural activit y. The appearance of the result is de termined by the choice of analys is method (mod- el) rather than by the struc ture of ac tive areas in the brain.

3. Read ing

3.1. Sequence of acti vation

It is usuall y assume d that when we see a familiar word basic visual featu res must be process ed first be fore the an al- ysis can proceed to the content , app arently first at the level of single letters and then as a whol e word whi ch furt her acti- vates the word’ s meani ng and its so und form. Accor ding to the influenti al dual-ro ute model ( Colthear t e t al., 1993 ) unfa- miliar words or nonw ords can not be ha ndled by this lexical route but, inst ead, we process them letter-by -letter, co nvert- ing each graphe me to its corresp onding phone me, thus resulting in a phonolo gical repres entation for the letter- string whi ch may or may not evo ke semant ic associa tions. Both ro utes are though t t o b e activated but the dom inance between the rou tes varie s with the familiarit y o f the words. An altern ative account is the distribut ed connec tionist mod- el (Pl aut et al., 1996; Seidenb erg and McClell and, 1989 ). Here, one assumes that both familiar and unfamili ar words are han dled by exactl y the same network, where orthogr a- phy, phonol ogy, and semant ics are pro cessed simulta neous- ly in a single strong ly inter active process , an d it is the amount of exposure to letter- strings rather than their lexi cal status that influ ences the system be haviour. The existing theoret ical models of reading are ba sed large ly on analys is of behaviour- al react ion times and error types in acquire d and develop- mental reading disorde rs.

Fi g. 5 summa rizes the corti cal dynami cs of silent read- ing, as reveal ed by MEG. First , activati on reflect ing basic visual featu re an alysis is detect ed around the oc cipital mid- line, at ab out 100 ms. It is followe d by left-l ateraliz ed acti- vation of the occipitotem poral cortex at 150 ms, associ ated with letter- string analys is. This trans ient acti vation seems to indica te the first stage of language -speci fic process ing or, more general ly, categor y-spec ific analysis ( Tarkiain en et al., 1999, 2002 ). It is taken to reflect pre-lexica l pro cess- ing as the response does not di erentiate between words and nonwo rds, or even conson ant strings ( Corne lissen et al., 2003b; Sa lmelin et al., 1996; Wydell et al., 2003 ).

ARTICLE IN PRESS

Lippincott Williams & Wilkins 2000

R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx

Visual feature

analysis

Neurophysiology xxx (2006) xxx–xxx Visual feature analysis Non-specific 0 400 800 Letter-string analysis Words =
Non-specific 0 400 800
Non-specific
0
400
800

Letter-string

analysis

Words = Nonwords 0 400 800
Words =
Nonwords
0
400
800

Time (ms)

Lexical-semantic

analysis

Nonwords 0 400 800
Nonwords
0
400
800

5

Fig. 5. Cortical dynamics of silent reading. Dots represent centres of active cortical patches collected from individual subjects. The curves displ ay the mean time course of activation in the depicted source areas. Visual feature analysis in the occipital cortex ( 100 ms) is stimulus non-specific. The stimulus content starts to matter by 150 ms when activation reflecting letter-string analysis is observed in the left occipitotemporal cortex. Subsequent activation of the left superior temporal cortex at 200–600 ms reflects lexical-semantic analysis and, probably, also phonological analysis. Modified from Salmelin et al. (2000a).

Readi ng co mprehens ion is reflect ed in the sub sequent acti - vatio n o f the left superi or tempor al cortex at 200–6 00 ms ( Halgren et al., 2002; Helenius et al., 1998; Pyl kka¨ nen et al., 2002, 2006; Pyl kka¨ nen and M arantz, 2003; Simos et al., 1997 ). This susta ined acti vation di erent iates betwe en words and nonwo rds (Sa lmelin et al., 1996; Wil- son et al., 2005; Wyd ell et al., 2003 ). Apart from lexical- se- manti c aspect s it also seems to be sensi tive to phonol ogical manipul ation ( Wydell et al., 2003 ). As discussed above, in speech percep tion ac tivation is conce ntrated to a rather smal l area in the brain an d w e have to rely on tim e informat ion to dissociate betw een dif- ferent pro cesses. Here, the di erent process es are sep arable both in timing and location. Becau se of that, one might think that it is easie r t o characteriz e language -related pr o- cesses in the visual than au ditory mo dality. Howeve r, he re the di cul ties appear at another level . I n reading , activ a- tion is de tected bilat erally in the occipi tal c ortex, along the tempor al lobe s, in the parietal cortex and, in v ocalized reading , also in the front al lobes, at various tim es with respect to stimulus ons et. Inter indivi dual varia bility furth er compli cates the pictur e, resulting in practicall y excessive amoun ts of tempor al and sp atial infor mation. The areas and time wi ndows de picted in Fig. 5 , with specific roles in reading , form a lim ited subset of all active areas observed during reading . In order to perfor m proper func- tional local ization one ne eds to vary the stimuli and tasks systemat ical ly, in a parame tric fashi on. Let us now consid- er how one may extract acti vation reflect ing pre-lex ical let- ter-str ing analys is and lexi cal-semant ic process ing.

3.2. Pre-l exical analys is

In order to tease apart early pr e-lexical pr ocesses in reading , Tar kiaine n an d colleagues ( Tarkiai nen et al., 1999 ) used words, syllables , and singl e letters, imbedded

in a noisy backgroun d, at four di eren t noise levels (Fig. 6 ). For control, the sequen ces also co ntained symbol strings. One sequen ce was composed of plain noise stimuli. The stimuli were thus va ried along two major dimension s:

the amount of featu res to pro cess increa sed wi th noise and with the numb er of items, letters or symbol s. On the other hand, word-lik eness was highest for clearly visible co mplete words and lowest for symbol s and noise. At the level of the brain, as illustrated in Fig. 7 , the data showe d a clear dissoc iation be tween tw o process es within the first 200 ms: visual feat ure analys is occurred at about 100 ms after stimulus present ation, wi th the active areas around the oc cipital midline, along the ventra l stream. In these areas, the signal increased with increa sing noise an d with the num ber of item s i n the string, sim ilarly for letters and symbol s. Only 50 ms later, at about 1 50 ms, the left inferior occipi totemp oral cortex showe d letter- string spe- cific activati on. This signal increased with the visibil ity of the lett er strin gs. It was strong est for words, weake r for syl- lables, a n d sti ll weake r for singl e letters . Cruc ially, the acti- vation was significan tly strong er for letter than symbol strings of equal length . Bilateral occipitotem poral activati on at abou t 200 ms post-sti mulus is consis tently report ed in M E G studies of reading ( Corne lissen et al., 2003b ; Pammer et al., 2004; Sa l- melin et al., 1996 , 2000b ) but, inter estingly, functi onal specificity for letter- strings is fou nd most systemat ically in the left hemis phere. The M EG d a t a on letter- string spe- cific activati on are in goo d agreem ent with intracran ial recordi ngs, bot h with respect to timing and location an d the pre -lexical nature of the acti vatio n (Nob re et al., 1994 ).

3.3. Lexical -semant ic analys is

To identi fy c ortical dynami cs of reading comprehens ion, Helen ius an d colle agues ( Helen ius e t al., 1998 ) employ ed a

ARTICLE IN PRESS

Oxford University Press 1999

6

R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx

R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx Noise Letter/ Syllable/ Word/ 1 symbol 2
R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx Noise Letter/ Syllable/ Word/ 1 symbol 2
R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx Noise Letter/ Syllable/ Word/ 1 symbol 2
R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx Noise Letter/ Syllable/ Word/ 1 symbol 2
R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx Noise Letter/ Syllable/ Word/ 1 symbol 2
R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx Noise Letter/ Syllable/ Word/ 1 symbol 2
R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx Noise Letter/ Syllable/ Word/ 1 symbol 2
R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx Noise Letter/ Syllable/ Word/ 1 symbol 2
R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx Noise Letter/ Syllable/ Word/ 1 symbol 2
R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx Noise Letter/ Syllable/ Word/ 1 symbol 2
R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx Noise Letter/ Syllable/ Word/ 1 symbol 2
R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx Noise Letter/ Syllable/ Word/ 1 symbol 2
R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx Noise Letter/ Syllable/ Word/ 1 symbol 2
R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx Noise Letter/ Syllable/ Word/ 1 symbol 2
R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx Noise Letter/ Syllable/ Word/ 1 symbol 2
R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx Noise Letter/ Syllable/ Word/ 1 symbol 2
R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx Noise Letter/ Syllable/ Word/ 1 symbol 2
R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx Noise Letter/ Syllable/ Word/ 1 symbol 2
R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx Noise Letter/ Syllable/ Word/ 1 symbol 2

Noise

Letter/

Syllable/

Word/

1 symbol

2 symbols

4 symbols

Fig. 6. Paradigm for focusing on pre-lexical processes in reading. The amount of features to analyze (four levels of noise) and word-likeness of the st imuli (symbols; letters, syllables, words in Finnish) were varied parametrically. Modified from Tarkiainen et al. (1999).

Visual feature analysis ~100 ms Letter-string analysis ~150 ms = >
Visual feature analysis
~100 ms
Letter-string analysis
~150 ms
=
>

Fig. 7. Dissociation of basic visual feature and letter-string analysis within 200 ms post-stimulus. Dots represent centres of active cortical patc hes collected from the individual subjects. Arrows indicate increasing strength of activation. Amount of features to analyze influenced activation strength at th e first stage ( 100 ms) and content at the next stage ( 150 ms).

visual version of sentenc es that creat e a very high expecta - tion for a certain fina l word (see section on speech percep - tion ab ove). In these sentenc es, the fina l word was either the exp ected one or one of three types of unexpe cted end - ings: totally anomal ous, une xpected but wi th an accep table meani ng in the sentenc e c o ntext, or semantical ly wrong but (decept ively) sharing its initial letters with the expecte d word. Fig. 8 displ ays the average time co urse of activati on for the sentenc e-endi ng words, with the left -hemisphere N400m activati on grad ed by semant ic congruit y. The response was strong est and lasted longest for the semant i- cally wrong word types and smal ler a n d shorte r-lastin g for unexp ected words with accepta ble meani ng. To the expect- ed words there was no respon se that would have exceeded the noise level . In the right he misphe re, abou t half of the subject s sho wed a qua litatively similar susta ined response but with much weake r di eren tiation between stimulus types. In reading , the ne ural signat ures of lexica l-semant ic process ing thus seem to be fairly strong ly laterali zed to the left hemisph ere. When the active areas are mode lled as focal ECDs, sources of the N400m response are con sistently localized to the sup erior tempor al co rtex ( Halgren et al., 20 02; Hel-

Expected

Meaning

Letters

Anomalous

Activation strength

Meaning Letters Anomalous Activation strength (nAm) 10 0 200 400 600 Time (ms) Fig. 8. Neural
(nAm) 10 0 200 400 600
(nAm)
10
0
200
400
600

Time (ms)

Fig. 8. Neural correlates of lexical-semantic processing in reading. The curves display the mean time course of activation in the left superior temporal cortex. The semantically wrong endings (totally anomalous, same initial letters as in the expected word) resulted in a prominent N400m response. The signal was smaller for the equally unexpected words that had a plausible meaning in the sentence context, and essentially non- existent for the expected words.

enius et al., 1998; Pyl kka¨ nen and Marant z, 2003; Pylkka¨ - nen et al., 2002; Salme lin et al., 1996; Simos et al., 1997 ), in the imm ediate vici nity of the auditor y cortex (Hele nius et al., 1998 ). Distri buted source mod els suggest furt her spreadin g of acti vation to the an terior tempor al and infer i-

ARTICLE IN PRESS

R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx

7

or front al cortex ( Halg ren et al., 2002; Marinkovi c et al., 2003 ). Involve ment of the left tempor al pole in semant ic process ing would agree with previous intracrani al reco rd- ings ( Halgren et al., 1994; Nobre an d McCa rthy, 1995 ) which , howeve r, did not pr obe the superio r tempor al cortex .

4. Spe ech product ion

Rese arch into the neural basis of langua ge pr oduction is compli cated by the strong artef act signal s that are generat - ed by mou th and tongue move ment and mask the corti cal activit y. For tunately, those distu rbing field patterns can often be remove d from the M EG data (Salme lin et al., 1994, 20 00b ). Norm ally, there is consider able intertrial var- iabilit y in tim ing from the tri gger stimulus to actual speech produ ction, and also from onset of mouth movem ent to actual speech onset (100–2 00 ms). This jitter makes it pos - sible to dissoc iate the artef act signal from the co rtical acti v- ity of interest . Let us consider two experi menta l setups that encomp ass speech prod uction, vocalized reading and pic- ture na ming.

4.1. Read ing aloud

Fig. 9 displ ays a sequence of acti vation dur ing v ocalized reading , colle cted from a group of 10 subject s ( Salme lin et al., 2000b ). Isola ted words (Ger man nouns composed of 7–8 lett ers) were present ed for 300 ms. After a delay of 500 ms a questio n mark appeared for 2 s , pro mpting the subject to read the word out loud . The oc cipital an d the left and right infer ior occipi totempo ral cortices were active within the first 200 ms. Based on the neural sequen ce in covert reading , out lined above, the early response c lose

to the occipi tal midl ine is likel y t o reflect , at least partly, the visual feat ure analys is, whi ch is followe d b y letter- string specific process ing in the left inferior occipitotem po- ral cortex . The large ly visual nature of the right inferior occipi totemporal signal is emphasize d b y the secon d response to the que stion mark. Act ivation in the left supe- rior tempor al and inferior parietal corti ces, star ting at about 200–300 ms afte r word onset, and reachi ng the max- imum at about 400 ms probably refl ects semantic pro cess- ing, as suggested by studi es on silent reading (N40 0m). Activat ion of the left inferior front al cortex , approxim ately Broca’s area, howeve r, is typic ally not observed in silent reading . The activit y star ted at ab out 200 ms afte r word onset, and is likely to reflect access to phonolo gical repre- sentation of the word for arti culation ( Fiez and Peters en, 1998 ). Act ivation of this region seems to be specific to vocalized reading . The responses depict ed in the two left-mo st columns fade out be fore the vocali zation pr ompt. How ever, the acti- vation s d e picted in the right-m ost co lumn begin at about 200–300 ms afte r word onset and persi st until actual vo cal- ization and even be yond it. This seems reason able as they arise from the left an d right sensori motor and premot or cortices an d apparen tly from the supp lementary motor area. Event-re lated modulat ion of rh ythmic cortical activit y i s less sensi tive to dist urbances caused by mou th movem en ts than phase-lo cked evoked respon ses and may, thus, pro- vide a useful tool for evaluating motor cortex invo lvement in speech producti on. The hea lthy human bra in typicall y shows parieto –occipital rhythm ic acti vity in the 10-H z range, the a rhythm , a n d roland ic activit y c o ntaining both 10- and 20-Hz componen ts, the l rhythm ( Hari and Salme- lin, 1997 ). The a rhythm is suppress ed by open ing the e yes,

10 fluent speakers speech word ? 0 300 800 ms 10 nAm 0 800 0
10 fluent speakers
speech
word
?
0
300
800
ms
10 nAm
0
800
0
800
0
800

Time (ms)

Fig. 9. Cortical dynamics of vocalized reading. Average cortical sequence of activation in 10 fluent speakers when they were reading isolated words ou t loud. The dots show the centres of the active cortical patches collected from the di erent subjects, and the curves the average time course of activati on in those areas. The first vertical line indicates the word presentation and the second vertical line the onset of the question mark that served as vocaliza tion prompt.

ARTICLE IN PRESS

8

R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx

with little e ec t o n the rolandi c activit y, an d the l rhy thm by moving the left or right ha nd. In the resting br ain, the sources of l rhythm concentra te in and aroun d the hand repres entation area in the centra l sulcus . The 10-H z c om- pone nt origi nates largely in the somato sensory cortex, but also prece ntrally, whereas the 20-Hz componen t seems to be predomi nantly a m o tor corti cal rhythm ( Salmelin and Hari, 1994; Salme lin et al., 1995 ). Importa ntly, for volunta ry movem ents the motor co rtex 20-H z activit y shows somat otopic organiz ation ( Salme lin et al., 1995 ), which means that one may use 20-H z acti vity to evaluat e function ality of di erent parts of the motor cor- tex, not only the hand area. In speech pr oduction, sup pres- sion of the 20 -Hz acti vity, taken as a signat ure of cortic al involv ement in task perfor mance, is detected in the mouth area bilat erally. In vocaliz ed reading of isol ated words the mouth- area 20-H z activit y i s sup presse d wel l before vocal- ization (Sa lmelin et al., 200 0b ). It turns out that this sup- pressio n i s correl ated with timing of the visual instructi on rather than mouth movem ent onset, clearly pointi ng to involv ement of motor cortex in cognit ive pro cessing, not simp ly motor co ntrol ( Saarine n e t al., 2006 ). These mea- sures indicate no clear speech- specific late ralizati on at the level of the mouth motor cortex . Interesti ngly, the hand areas ap pear to be involv ed in speech producti on a s well althoug h obviousl y much less than the mouth a reas an d only during the actual move ment ( Salmelin et al., 2000b ). When the linguist ic compo nent of the output was varied from nonv erbal kissing mov ement to silent artic ulation of the v owel /o/ and furt her to repeat ed vocali zation of the same word and generat ion of new words the relat ive contrib utions of the hand and mouth areas were a ec ted in a systemat ic fashi on ( Salme lin and Sa ms, 2002 ). In the mouth area, the 20-Hz sup pression remai ned essent ially the same for all tasks but in the hand area the suppress ion was dimi nished with increa sing lingui stic demands of the task. Accor dingly , the mou th vs. han d area segreg ation in the motor co rtex was strong er for verbal than nonv erbal mou th movem en ts ( Saarine n e t al., 2006 ; Salme lin an d Sams, 2002 ), an d this e ect seems to be ind e- pend ent of seque nce length or co mplexity (Sa arinen et al., 2006 ). The main facto r appears to be the verbal vs. nonv er- bal contrast, that is, the hand areas are mo re involv ed when pr oducing nonverba l than verbal mouth movem ents.

4.2. Pictur e nami ng

Object na ming is a ba sic functi on of langu age that is thought to include all the main stage s of word produ ction from con ceptual ization to selec tion of a lexi cal item , pho- nologic al en coding and preparat ion for articu lation (Lev elt et al., 1999 ). Due to its app arent sim plicity it is a frequent ly used task in clinical practi ce. However, from the neurosci - ence point of view the proce ss of picture naming is far from simp le. Unlike for reading and speech percep tion, it has been di cult to associ ate specific tim e wind ows a n d brain areas with process ing stage s assum ed to be involv ed in pic-

ture naming . A t present, there is only a smal l numb er of MEG studi es describ ing the sequence of acti vation during overt pictur e naming ( Levelt et al., 1998; Salmelin et al., 1994; So¨ ro¨ s e t al., 2003 ). These studies give only indir ect evidence on the timing and locat ion of the cognit ive sub- process es of word prod uction, as experi menta l manipul a- tions potentiall y di erent iating the subprocess es wer e either not used ( Salme lin et al., 1994; So¨ ro¨ s e t al., 2003 ) or failed to reveal di erence s i n acti vation ( Levelt et al., 1998 ). Naming a set of pictur es which were semant ically related or not suggest ed prim ing e ects in the left tempor al cortex at arou nd 2 0 0 m s post-sti mulus , interprete d as a sig- nature of lexical retr ieval ( Maess et a l., 2002 ). In defrey and Lev elt (2004) have compil ed timing infor - mation obtaine d from behavior al and event-rela ted poten- tial studi es (EEG) with sp atial infor matio n provided by neuroim aging methods , mainl y posit ron emission tomogr a- phy (PET) an d functio nal magn etic resonance imag ing (fMRI), includi ng data from the smal l number of M E G studies a vailable at presen t. The meta -anal ysis suggest s the followin g approxim ate time wind ows and neuronal cor- relates for the sub process es: (i) visual object recogni tion and co nceptual ization at 0–175 ms post stimulus , involv ing occipital and ventrot emporal regions ; (ii) selection of the corresp onding semantic- syntac tic represen tation (a lemma) from the menta l lexico n at 175–250 ms, associ ated with the mid-sect ion of the left middle tempor al gyrus; (ii i) phono- logical code retr ieval at 250–330 ms, involving pos terior parts of left middle and superio r tempor al gyri (i.e., Wer- nicke’s area) ; and (iv) preparation for or al outp ut after 330 ms, engaging Br oca’s area in the left infer ior front al gyrus and bilateral sensori motor areas. Fur ther resear ch combini ng MEG an d caref ul parame tric varia tion of tasks and sti muli is clear ly needed to verif y or correct this view. As illu strated in Fig. 10 , the context of pictur e naming may influence the observed pattern of acti vation . W h e n naming stand- alone pictures of object s ( Leve lt et al., 1998; Salmelin et al., 1994 ) activati on pro ceeded from the occipital co rtex (<200 ms) to bot h tempor al and parietal areas (>200 ms) an d furt her to infer ior front al cortex (>300–40 0 ms) bilateral ly. Ho wever, when naming object s or acti ons from sim ple draw ings of events ( So¨ ro¨ s et al., 2003 ) activati on followe d a route from oc cipital cortex (<200 ms) through the parieto –occipito–t emporal juncti on (>200 ms) to the left dorsal motor /premot or cortex , i n par- ticular (>400 ms). The e ect of stimulus and task on the neural correl ates of overt na ming needs to be address ed in futur e studi es.

5. Pre-su rgical mapping of langua ge function

5.1. Language- dominan t hemispher e

Bas ed on the above, ea rly stages of speech percept ion and analys is of wri tten language would seem to sho w the clearest left-hem ispher e lateralizati on in healt hy right- handed indivi duals. Lateral izat ion of the N100m response

ARTICLE IN PRESS

R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx

9

a
a
b
b
/ Clinical Neurophysiology xxx (2006) xxx–xxx 9 a b Fig. 10. Cortical activation in picture naming.
/ Clinical Neurophysiology xxx (2006) xxx–xxx 9 a b Fig. 10. Cortical activation in picture naming.

Fig. 10. Cortical activation in picture naming. (a) Activated areas when naming pictured objects. Solid and dashed lines indicate distribution of active areas as reported in Salmelin et al. (1994) and Levelt et al. (1998), respectively. (b) Activated areas when naming objects or actions from simple drawings of events as reported in So¨ ro¨ s e t al. (2003).

to speech vs. non-speech sou nds indeed shows promi se for

a clini cal paradig m, as it is fast and ea sy to perfor m and simp le to analyze . Probably the simplest approach was pr o- posed by Goo tjes and colle agues ( Gootj es e t al., 1999 ) who present ed healt hy right -handed sub jects with pairs of tones,

pairs of piano notes, and pairs of vo wels in a random sequen ce. The subject s’ task was to detect targe t pairs in which the tw o sounds wer e identical. The authors used no source mode lling but c ompared the respo nses to speech vs. non-sp eech so unds by selec ting, over each hemisp here, the sensor that sh owed the strong est N100m respo nse; impor tantly, the MEG da ta wer e recorded using planar gradiom eters that detect the maximu m signal direct ly above an active cortical area. The relative stren gth of the N100m response to vowels vs. tones was consis tently stron- ger in the left than right hemisp here in the right -handed indivi duals. Kirveskari and colleagu es ( Kirveskari et al., 2006 ) furt her evaluat ed this ap proach by sim plifying the paradig m t o only contain vowel s and tones ( Fig. 1 1 ) and includi ng both right - a n d left-handed subjects. The au thors demon strated a di erence in hemis pheric dominance betwe en right-hand ed and left-handed sub jects that closely

resem bles the ratio of left- vs. right -hemisph ere lateraliza - tion of languag e functi on suggest ed by previous invasive studi es as well as by an atomical an d fun ctional compari - sons between left - and right -handed indivi duals. This sim - ple pa radigm promi ses to be a reasonabl e candidat e for compari son with the Wada test, and eventual clinical use. The co rrespondenc e betw een MEG measur es and the Wada test ha s b e en explore d i n a numb er of studi es. Papa - nicola ou and colleagues have use d a receptive language task, de tection of repeat ed words, both in the au ditory and visual moda lity. They focus on the late susta ined acti - vatio n (>200 ms) and model the distribut ion of a ctivity with an ECD every 4 ms, sep arately in each hemisph ere.

A

late ralization index is e stimated from the total numbe r

of

ECDs in the left vs. right hemis phere that pass a set of

accepta nce crit eria develop ed over a series of related exper-

iments (Br eier et al., 19 99a, 2000; Simos e t al., 1998; Zou- ridakis et al., 1998 ). The authors report a high degree of concordan ce (>85%) be tween their M E G measur e of late r- alizati on and the Wad a test (Breier et al., 1999b, 2001; Papan icolaou et al., 2004 ). Apart from the mult iple studies on English -speakin g subject s perfor med by Papa nicola ou and colleag ues the same pa radigm and analys is app roach have been test ed on Spa nish- speaking subjects, with com- parable resul ts (agre ement be tween MEG evaluat ion an d the Wad a test in 7 of 8 subject s; Maest u e t al., 2002 ). A simila r a n alysis method has also been used to de monst rate lateralizati on in pe rception of vowels vs. tones ( Szymans ki et al., 1999 ). Accor dingly , from the clinical point of view, the app roach seems pro mising a lthough furt her tests by other grou ps are clearly needed. It is worth not ing, howeve r, that from the neurosci ence point of view it is not immediat ely obvious why the number of accepta ble ECDs sho uld be a relev ant or useful measur e of lateralizati on (ins tead of the stren gth of the activati on or its overal l time co urse). For exampl e, if two areas that are fairly close to each other but spati ally separabl e are active simulta neously (see, e.g. , Hele nius et al., 1999a ) one woul d not be able to accou nt for their common patte rn of activa- tion with a singl e ECD that woul d pass any reasonab le cri- teria of accepta nce, which would pot entially result in an erroneous de scription of laterality. Never thele ss, as Papa- nicola ou and co lleagues focused on the tim e window >200 ms afte r stimulus onset a n d foun d the ECDs locat ed essential ly in the pe risylvian region they seem to be focu s- ing on the sustained N400m response. In many subjects, the num ber of accepta ble ECDs (salient dipolar field pa t- terns) may well serve as an indir ect, app roximate measur e of the strength and durati on of the N400m activati on. Strength of cortical activati on in the left vs. right hemi- sphere ha s also be en co mpared wi th the Wada test. Bowyer and colle agues (Bow yer et al., 2005 ) used a current de nsity imaging techni que (MR-FOC USS) to local ize ne ural acti- vation during tw o language tasks, verb generation and pic- ture na ming, that wer e performed menta lly, withou t overt vocalizati on. Hirata and colleague s ( Hir ata et al., 2004 ) localized task-rel ated suppress ion on rhythm ic activit y i n the b (13–25 Hz) an d low c (25–50 Hz) band during a silent reading task, using Syn thetic Aperture Magnetom etry (SAM ). In both studi es, source analys is suggest ed involve- ment of multiple brain areas, includi ng the pos terior tem- poral (approxi mately Wernicke ’s area), inferior parieta l, and inferior front al cortex (approxi mately Broca’s area). Lateralized activa tion of the inferior frontal cortex , a s mea- sured by increased current density of phase-l ocked evo ked responses ( Bowyer et a l., 2005 ) o r suppress ion of rhythmic activit y ( Hirata et al., 2004 ), was found to be in best ag ree- ment with the resul ts of the Wada test (>90%) . Left-later- alized acti vation in Broca’s area (and Wernicke ’s area) in right-hand ed individ uals ha s also been report ed by Kober and c olleagues (Kob er et al., 2001 ) who used a spatial fil- tering method to study acti vation of poster ior tempor al and infer ior front al cortex during silent reading and pictur e

ARTICLE IN PRESS

10

R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx Signal Vowels Signal strength strength Tones (fT/cm)
R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx
Signal
Vowels
Signal
strength
strength
Tones
(fT/cm)
(fT/cm)
200
200
0
200
400
600
0
200
400
600
Time (ms)
Time (ms)

Fig. 11. Auditory responses to vowels vs. tones in one subject. The MEG measurement helmet is viewed from above, flattened to the plane. The curves show the variation of magnetic field as a function of time. This MEG system (Vectorview , Elekta-Neuromag Ltd.) has planar gradiometers that detect the maximum signal directly above an active cortical area. At each measurement location there are two orthogonally oriented planar gradiometers. Th e rectangles over each hemisphere indicate the sensor pair that detected the strongest N100m response. The inserts depict the overall signal strength at those sites (vector sum, i.e., square root of the sum of the squared signal strength in the two orthogonal sensors). The N100m response in the left hemisphere is markedly stronger to vowels than tones whereas, in the right hemisphere, there is no clear di erence. The analysis was performed by computing the mean signal strength in a 50-ms time window centred on the N100m peak latency (grey bar), and the ratio of signal strength for vowels vs. tones, separately in each hemisphere. The laterality index is obtained as the di erence between left- and right-hemisphere ratios divided by their sum. The absolute valu es of the signal strength are thus irrelevant. Modified from Kirveskari et al. (2006).

naming in he althy sub jects. Again , these techn iques hold promi se but more studi es by di erent grou ps are ne eded to evaluat e their practical feasibility an d reli ability.

5.2. Mappin g activation in langua ge tasks

A c o mprehens ive mapping of areas involv ed in language process ing be comes a relev ant issue when the presence of a tumor or an epileptic focus necessi tates ope ration on the language -domi nant he misph ere and, more specifica lly, when the resect ion is likel y to border on areas typic ally thought to be involv ed in language process ing. The need for pre-su rgical mapping – a nd post- surgical evaluat ion – of language areas is stre ngthened furth er if language defi- cits emerg e i n everyday inter action or if they are reveal ed in ne uropsychol ogical test ing. Ideally , a n MEG charact er- ization of language func tion could help to plan and speed up the evaluation perfor med with direct cortical sti mula- tion dur ing actual surger y. Providi ng a map of the language functio n for the neu ro- surgeon is far more problem atic, bot h techni call y and con - ceptual ly, than local izing , e.g., the hand , foot, and mouth repres entation areas in the somatosen sory an d motor co r- tex ( Ma¨ kela¨ et al., 2001 ). Ther e are severa l aspect s to lan- guage functi on whi ch may involv e partl y overla pping but also quite di erent co rtical areas. Thi s may be because

the neural networks involv ed are di erent or because dis- tinct parts of the same general network are emphasi zed in

di erent tasks. Per ception may occu r via auditor y, visual,

or tactile pathw ays. Com prehen sion of single words, sen-

tences, or continuous text or discou rse may rely on sligh tly

di erent netwo rks. Semant ic and syntac tic pr ocessing are

likely to have partly separat e co rtical repres entat ions. Also,

language recep tion and prod uction are two (partly) di er-

ent things. Produ ction c an happen via speech , writi ng, or signing. The c orrect prosod y, the rhythm and intonat ion,

is a relev ant aspect of speech producti on and co mprehen-

sion. It is not imm ediat ely clear whi ch aspect of language

should be mapp ed for clinical pur poses. Thi s process shou ld probably be guided by a thorou gh neuropsycho logical identifica tion of potential problem s induced by tumor /epilepsy in specific aspect s o f language , say, speech compreh ension. A targeted neurosci ence-bas ed parame tric design could then be used for functio nal local-

ization of brain areas involv ed in that pa rticular process ing stage (see abo ve). Anothe r pos sibility woul d b e t o develop

a set of recept ive an d exp ressive language tasks that woul d

be as general and realist ic as possible to maxi mally engage

the relevan t neural netwo rks which could then be test ed with direct cortical sti mulation; these are not ne cessarily experimen tal designs one woul d typic ally use in basic (non-clini cal) resear ch of language function .

ARTICLE IN PRESS

R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx

11

At presen t, there are few publis hed studies to add ress these issues. Simos and colle agues ( Simos et al., 1999 ) report ed that cortical areas activ ated during visual and auditor y word recogn ition tasks , identified as cluste rs of ECDs fitted to the left- and right -hemisph ere MEG pat- terns every 4 m s (see abov e for paradig ms used to evaluate laterali zation ), were in good agreem ent with tho se de tected during direct cortical stimula tion. Kama da and colleagu es ( Kama da et al., 2004 ) tracke d ne ural correlates of letter percept ion in a patie nt who unde rwent a radica l resection of a mesi al tempor al gliom a, involv ing the left fusi form gyrus. As anticipat ed by preop erative MEG mapping of letter percept ion the resect ion init ially resul ted in severe dyslex ia, appa rently due to the loss of the area specia lized in early letter-str ing analys is (see abo ve for secti on on read- ing). Duri ng a one -year follow-u p the subject ’s reading skills wer e gradual ly impr oved. MEG data indica ted con - current ov ershoot of activit y in the spared left sup erior tempor al cortex (late letter- string analys is, lexi cal-semant ic process ing) but no compensat ory acti vity in the right fusi - form gyru s. Patarai a an d colleagues ( Pataraia et al., 2005 ) local ized acti vation relat ed to aud itory word recogn ition/ repeti tion, with focus on Wernicke ’s area, be fore and afte r left anterior tempor al lobect omy in patie nts su eri ng from intrac table tempor al lobe epilep sy. The authors compared the M EG findi ngs wi th neurop sycholog ical assessment and found that postop erative increase of right -hemisphe re language acti vation was more likel y t o occur in patie nts who lacked clear hemisp heric laterali zation preoperat ively than in patients who showed left -hemisphere dominanc e. Spatia l distribut ion of language -rel ated activati on may be radica lly alte red in epilepsy ( Breier et al., 2005; Pata raia et al., 2004 ).

6. Lan guage d isorders

Published MEG research into the neural basis of devel- opmen tal and acquire d languag e disorde rs mostly falls in the realm of ba sic neurosci ence rather than in the clinical domain. W hile there is genu ine interest in unde rstanding the neu ral unde rpinni ngs of these disorde rs, as a prelude to possible ne uroscien ce-driven inter vention s, these deficit s also provide e ssential infor mation ab out br ain areas and time wind ows that may be pa rticular ly relev ant to success - ful language percep tion and prod uction. Comparis ons of cerebra l implement ation of language in sub jects who have functi onal disorde rs in specific aspect s of language witho ut obviou s structural defic its, such as dy slexia (readi ng) and stutteri ng (speec h produc tion), and in subject s with unim- paired language functi on have bee n highly infor mative in that regard. Lang uage disorde rs are e ci ently diagn osed and charac- terized be havioura lly. How ever, a detai led neuropsycho - logic al profile combined with a detai led descri ption of language process ing at the neural level in individ ual sub- jects could be a power ful tool. Even tually, it cou ld help to identify specific su bgroups of subjects within groups that

are now pooled unde r a single label acco rding to behav- ioural crit eria or lesio n site, facilitate developm ent of meth- ods for trai ning or rehabilit ation that woul d b e optima lly targeted for each indivi dual, and provide object ive means for evaluat ing the e cacy of a treat ment.

6.1. Dyslexia

MEG studi es on adult subject s have demonst rated that in dyslex ic individ uals visual feature process ing is normal but reading is disrupted at the subsequent letter- string spe- cific stage ( Salme lin et al., 1996; Hele nius et al., 1999b; Sa l- melin and Helen ius, 2 004 ). The non-exi stent or ab normally weak acti vation of the left infer ior occipi totem poral cortex in dy slexia has been co rroborate d b y later fMR I studi es (Paul esu et al., 2001; Shaywitz et a l., 1998 ). The subsequent left superi or tempor al activati on, reflect ing reading com- prehensi on, is weaker and delayed in dy slexic individuals (delay 100 ms; Hele nius et al., 1999a ). In fluent readers , the letter- string-sp ecific activati on ( 150 ms) appears to be the gatew ay from visual to linguist ic analys is, a fast route that automa tically sets letter-str ings apart from other object s and faci litates fast reading . The lack of this ’fast route’ for written language is likely to be the immediat e reason for the manifest di cul ties in reading in dyslexia. The functional de ficit a t the level of categor y-spec ific oc cip- itotempora l acti vation ( Tarkiai nen et al., 2002 ) seems to be related to writte n language , i n p a rticular , a s process ing of faces displays a normal pattern of activati on in this area and time window (Tar kiainen et al., 2003 ). These di cul ties in learning to read are often though t t o derive from impaired pro cessing of phone mes (Bra dley and Bryant, 1983 ). MEG studi es ha ve shown that in dy slexic individu als the strength of the superi or tempor al activati on evoked by sp oken words diverge s from the normal pa ttern by abou t 100 ms afte r stimulus onset (Hele nius et al., 2002a, b; Parviai nen et al., 2005 ), i.e., in the time window when phone tic infor matio n i s extra cted (Par viainen et al., 2005 ). Dier ences in activati on stre ngth between dy slexic and control grou ps from 100 ms onwar ds have be en report- ed for process ing of non-speech soun ds as well ( Fisher et al., 2006; Nagar ajan et al., 1999; Par viainen et al., 2005; Renvall an d Hari , 2002, 2003 ). In speech pe rception, the stage of lexical- semantic process ing is delayed in dy s- lexic adults (50 ms; Helenius et al., 20 02b ) but clear ly less than in reading . Simos, Br eier and colleagues have studi ed the neural correlates of speech percept ion and reading ( Simos et al., 2000a, b ) i n d y slexic and non-readi ng-impair ed childr en, using the word reco gnition parad igm and analysis methods developed for evaluation of language late ralizati on (see above). They report increa sed involv ement of the right tempor oparietal region and decreas ed involvemen t o f the left tempor opa rietal region c ompared wi th the pa ttern observed in con trols. When ch ildren wer e asked to discr im- inate between speech soun ds (syllabl es from /ga/-/ka/ con- tinuum ) a n increa sed relat ive a ctivatio n of the right

ARTICLE IN PRESS

12

R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx

tempor oparieta l areas was fou nd to be correl ated with reduced pe rformance in phonol ogical tasks ( Breier et al., 2003 ). Comp arison of word reading an d pictur e naming tasks in dyslex ic vs. co ntrol childr en points to abnormal i- ties in neural process ing in dy slexic individ uals that are spe- cific for reading ( Trau zettel-Klos inski et al., 2006 ). In ad dition to a ctivatio n stre ngth or timing or number of accep ted ECDs , interhemi spheri c symm etry in the loca- tion of the active areas ha s also been report ed to di eren- tiate between g roups of dy slexic and non-read ing- impaired subjects. Norm ally, the centre of activati on dur - ing the au ditory N100m response is locat ed 5–10 mm more front ally along the suprat emporal plane in the right than left hemisp here (e.g ., Pantev et al., 1998; Sa lmelin et al., 1999 ). In dyslex ic subjects, howeve r, the N100m response in ad ults (Heim et al., 200 3a ) and the early auditor y acti va- tion in ch ildren ( Heim et al., 2003b ) did not show the expecte d spatial asymmetr y. The degree of the reduced hemis pheric asymm etry, suggest ed to be related to a n an a- tomical asymm etry of the planum tempor al, appeared to be correl ated with phonolo gical abiliti es (Paul et al., 2006 ). For these ea rly aud itory respon ses, the ne ural cu rrent flows perpend icular to the co urse of the sylvian fissure. As the spatial resol ution of M E G i s best in the plane orient ed orthogon ally to the direction of current flow, wher e small changes of sou rce location result in large changes in the magnet ic field pattern, the smal l di erence s i n source loca- tion along the sup ratempo ral plane may be relev ant. For most pur poses, howeve r, the pa rticular strength of MEG is in the power ful combinat ion of accu rate tim ing with approxim ate (1–2 cm) locat ion. In both reading an d speech percept ion cortical process - ing in dyslex ic indivi duals app arently starts to di er from the normal pattern at the earliest language -specifi c process - ing stage and there is a dramat ic di erence in tim ing by the stage of semant ic analys is, whi ch is particu larly empha sized in reading . These obse rvations would seem to point to impaired integratio n o f auditor y and visual informat ion in dy slexia. Bec ause the pro blems in dyslexia appear so ear- ly on in the co rtical sequ ence, it seems essent ial to study these neural process es in ch ildren who are in the process of learni ng to read. Parviain en and colleagues ( Parviaine n e t al., 2006 ) investiga ted neural correl ates of visual percept ion of let- ter-str ings in 7-year -old children who were on the first grade of the elem entary school, using a simplified versi on of the letters -in-no ise paradig m origi nally develop ed for adults (cf. Fig. 6 ). Based on a set of standar dized behav- ioural tests, these ch ildren were exp ected to beco me fluen t readers . I n children, the sequen ce of activatio n was func- tionally quite sim ilar to that in a dults, with visual feature analys is in the occipi tal cortex follo wed by lett er-string- specific a ctivatio n in the left occipi totempo ral cortex , and finally susta ined acti vation in the left su perior tempor al cortex . How ever, all stage s were delayed in tim e, by abou t 50 ms at the stage of visual feature an alysis and by about 100 ms at the stage of letter- string an alysis. A letter-

string-sp ecific response was detect ed in ab out half of the children , which is clear ly less than in fluently read ing adu lt subjects (>90%) . I n those childr en who did sho w letter- string-sp ecific activati on its stre ngth was correl ated with behavioura l measur es of phonol ogical awar eness. As pho- nologic al skills are seen as a prerequi site for reading acqu i- sition, this is an intrigui ng finding. Simos and colle agues (Simos et al., 2002b ) found, in Engl ish-speak ing ch ildren, that a risk for developi ng reading pro blems was accompa- nied by reduced engagem en t o f the left sup erior tempor al region an d increa sed acti vation of the co rrespondi ng right-hem ispher e region. The authors repo rted a nor mali- zation of this aberrant pattern followin g remedi al training (Simos et al., 2002a ). Ther e i s a n obvious ne ed for controlled studi es of the

e ects of inter venti on on read ing skills in children and adults that are based on choices of trai ning paradig ms informed by neuro science. A comb ination of behavioura l and neural measur es sho uld be used for eva luation of the results. The oc cipitotem poral letter-str ing activati on seems like one pro mising cand idate for this pur pose but more research is clearly needed into its developm ent and role in reading acquisiti on.

6.2. Stutteri ng

W hen fluent speakers read aloud isolated words the sequence initially follows that observed in silen t reading (see above). In add ition, acti vation is observed in the left inferior front al co rtex when prep aring to speak , pro bably reflecting access to the phonol ogical repres entat ion of the word for articulat ion, and in the left and right mo tor and premoto r cortex and supplem entary motor area during actual speech produ ction ( Salme lin et al., 2000b ). Intrigu- ingly, these are the very areas in whi ch the tim ing an d strength of acti vation in developm ental stuttere rs ha s been found to di er from that in control s, thus indica ting abnor- malities in process es specifica lly involv ed in ov ert spee ch producti on rather than core lingui stic analys is (Sa lmelin et al., 2000b ; Wall a e t al., 2004 ). Furtherm ore, a ctivatio n of these same areas was a ected also when stut terers simply listened to spoken sentenc es that they needed to repeat or transform after a short delay (Bi ermann-Rub en et al., 2005 ). Tr acking of anatom ical conn ectivity with MRI sup- ported the view of a defective inter play between the left inferior front al cortex and motor /premot or areas ( Sommer et al., 2002 ). Task-re lated suppress ion of the motor cortical 20-Hz rhythm during ov ert reading has further demon- strated a strong involv ement of not only the face but also the hand areas when stut terers prod uce speech (Sa lmelin et al., 2000b ), suggest ing that motor corti cal specia lization for verbal mouth movem ents may not have developed nor- mally in individu als who stut ter. A func tional de ficit in the interplay betw een auditor y and mo tor syst ems during speech producti on, suggest ed to unde rlie stutteri ng ( Fair- banks, 1 954 ), finds some support in MEG data ( Salmelin et al., 1998 ).

ARTICLE IN PRESS

R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx

13

A wealth of techni ques has been proposed to impr ove

fluen cy in stut terers ( Bloodste in, 1995 ). MEG should be an excell ent tool for evaluat ing the neural correl ates of flu- ency training but again, as in the case of dyslexia, it is essent ial to first gain a bette r unde rstand ing of the ne ural process es underlyi ng stut tering and of detailed neuro psy- chologi cal de scription of the training methods . Fr o m the neuro science poin t of view , stut tering is a somewh at pro b- lema tic pheno menon as it is typicall y not present when pr o- ducing isolated words. M ore realist ic situ ations of speech

produ ction, ideal ly con tinuous free discus sion, woul d be neede d to fully asses s stuttering . Suc h approac hes will requir e new techni ques for signal analys is that rely on ne u- ral timing rather than on stimulus tim ing.

6.3. Apha sia

MEG resear ch on aph asia, clinically a highly relevan t topic, is currently quite limited. The en tire language net- work may be drastical ly a ected by the lesio n. For exam- ple, a dissoc iation between naming acti ons (verbs ) and object s (no uns) ha s been repo rted in aphasia and thought to reflect di eren t cerebra l represen tations for acce ssing these lexi cal categor ies (Car amazza and Hilli s, 1991 ). Recent MEG da ta, howeve r, showe d a n essent ially identi - cal sequ ence of acti vation for ve rb and noun retrieval in healt hy subjects but clear diverg ence in an aph asic patie nt with specific di cul ties in naming nouns. It thus seems that dissoc iation of verb and noun prod uction at the cortical

level may only emerg e when the normal language network is disrup ted ( So¨ ro¨ s e t al., 2003 ); note, howeve r, that the questi on of noun vs. verb dissoc iatio n a t the neu ral level in normal subjects is far from settled, an d the resul ts also seem to influence d b y the task (e.g ., picture naming , lexi cal decisio n, semant ic categor ization of written words ).

In the presence of lesions one cannot sim ply assum e that

the neural responses displ ay the normal spatio tempora l distribut ion, and there is plenty of indivi dual varia tion in healt hy subjects as well. Therefor e, function al local ization is pa rticular ly impor tant in patie nt studi es. Fig. 12 shows an exa mple of such a proce dure in patie nt HH who had an extens ive lesio n i n the left hemisph ere, reachi ng to the left superi or tempor al cortex ( Laine et al., 2000 ). Based on neurop sycholog ical findings the lesion a ected his read- ing comprehen sion. M ore specifica lly, he sho wed a rare syndrom e know n as deep dy slexia wher e extens ive left- hemis phere damage results in specifica lly semantic errors in reading (e.g., ‘‘moo n’’ read as ‘‘cr escent’’). It ha s been though t that these errors refl ect shif ting of semant ic analy- sis to the right hemisph ere. Patient HH was tested using a set of semantical ly constrained sentenc es that en ded with expecte d o r une xpected words (see abo ve for secti on on reading ). The data demon strated that in HH lexical- seman- tic pro cessing was still subserv ed by the damaged left hemi- sphere as in unimpai red sub jects. Spatiot emporal patte rns of acti vation in a spoken word recogni tion task were compa red between six chron ic aph a-

sics and six control su bjects (Br eier et al., 2004 ). The data indica ted de crease d involv ement of areas typic ally associ at- ed with recept ive language functi on, includi ng the left supe- rior tempor al gyrus, but increa sed invo lvement of the surroundi ng a reas. The authors sugge sted that be tter recover y may be associ ated with normal izat ion of the pre- morbi d language areas rather than recrui tment of adjacent cortex . Languag e training in aphasia , its e cacy , and its co r- tical correlates are question s of major clinical impor - tance. Corne lissen an d colleagu es ( Corne lissen et al., 2003a ) investiga ted the neural correl ates of re-learni ng in three ano mic patien ts. Based on extens ive behaviou ral testing, these patie nts were assum ed to have weak seman- tic associati ons, thus resul ting in an input to the phono- logical process ing stage that was not coherent enough to produce the word. Duri ng trai ning, pictures wer e shown in matr ices which also co ntained other semantical ly relat- ed item s, e.g., a co w together with other domestic a ni- mals. Repeated trai ning of naming on this type of matrix structure is expecte d t o strengthen the connecti ons in the semant ic system. After training, single object s should be easie r t o name with the help of the strong er semantic network. In all three subject s, the left inferior parietal cortex was the only area sho wing e ects of train- ing. Behaviou ral improv ement was associated with increa sed corti cal activati on, ap proaching the level of activati on to initial ly easy-to-na me item s. Bas ed on loca- tion and tim ing, on the training procedure, and on the subjects’ beh avioura l pro file this acti vation was inter pret- ed to reflect more e ective phonolo gical enco ding of the target words . Ther e was no evidence of increa sed right- hemisp here participat ion afte r trai ning. In aphasic patien ts, the neural responses may show consider ably more intersess ion variab ility than in hea lthy control s. In ord er to reli ably evaluat e trai ning e ects it is impor tant to test the subject with the same paradig m a t least twice both before and afte r training ( Corne lissen et al., 2003a ). Presence of abno rmally ab undant slow rhythm ic activit y (<6 Hz), often detect ed in brain areas borderi ng a struc tural lesi on ( Butz et al., 2004; de Jo ngh et al., 2 003 ), also contrib utes to the app arent v ariability (Corne lissen et al., 2 003a ). On the other hand, the ab er- rant low -frequ ency activit y may also serve as a measur e of the e cacy of a training pro cedure. Meinzer and col- leagues ( Meinzer et al., 2004 ) found that the amou nt of perilesiona l low-freq uency activit y was correlated with the training- induced behavioura l impr ovement in lan- guage abili ties; it remai ned unc lear, howeve r, why the magni tude of rhythm ic acti vity was increa sed in about half of the patie nts and decreas ed in the other half. Accor dingly, it is feasib le to track an d qua ntify neural changes associ ated wi th re-learni ng of language a bilities in aphasia . A n e ssential step wi ll now be to ga in be tter underst anding of langu age learning in the health y brain in order to take full ad vantage of the pos sibilities in the clinical domain.

ARTICLE IN PRESS

MIT Press 2000

14

R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx

L R
L R
1 2
1 2
3
3

Sentence-final word

1

2

3

20 nAm Anomalous Expected
20 nAm Anomalous Expected

20 nAm

20 nAm Anomalous Expected

Anomalous

Expected

0

400

Time (ms)

800

Fig. 12. Characterization of lexical-semantic analysis in an aphasic patient. Patient HH had an extensive lesion in the left hemisphere, apparently encompassing the superior temporal cortex. Nevertheless, when tested with the sentence-reading paradigm designed to focus on lexical-semantic an alysis HH had two centres of activation (dots) in the left superior temporal cortex, with the anterior one showing a salient di erence between the responses to t he semantically anomalous and expected sentence-ending words, a pattern interpreted as reflecting lexical-semantic analysis and similar to that obse rved in healthy controls. No di erence between sentence types was detected in the right superior temporal cortex. Modified from Laine et al. (2000).

7. Conc lusion

A reasonabl e amo unt of infor matio n has begun to be availab le ab out the co rtical dynami cs of ba sic process es of speech percep tion, reading , and speech producti on to suppo rt clinical MEG studies of language function . For estimat ion of the language -do minant hemisp here mult iple experi menta l designs and analys is approach es have been propo sed whi ch will need to be furth er tested in practi ce. Ideally, the test shou ld be fast and easy for both the subject and the experi mente r, an d extra ction of the resul t should not requir e source estimat ion which takes tim e and may be influen ced, to some de gree, by the analysis method and cho ices made by the person perfor ming the analys is. One exa mple of such a paradig m was de scribed abo ve. When seeking to de scribe language repres entation in more detail experi menta l design is a key issue . Bot h function - al disorde rs and struc tural lesion s may severe ly alte r neu ral process ing. Functional local ization with caref ul parame tric varia tion of stimuli and/o r tasks is pa rticular ly impor tant in patien ts as one cannot sim ply assume that the respon ses displ ay the nor mal spatiotem poral dist ribut ion – and there is plenty of interind ividual varia tion in the healt hy popula- tion as well. Parametric designs tend to increa se the dura tion of the experi ments , thus emphasi zing the need for careful neurop sycholog ical testing to guide the MEG experimen ts for optim ally tapping the specific pro blems in language per- form ance of individ ual patie nts. If a grow ing tumor starts to degrade speech comprehen sion it woul d seem parti cularly relev ant to map that specific functio n in detail (function al mapping us ing a parame tric design) to guide the pos sible resection , o r decisio n about a resection, and combine it with an overvi ew provided by a nother, more g eneral language task that woul d encompass both recept ion and producti on (e.g., pictur e naming , vocalized reading ) without detailing specific functio nal su bcomponen ts. From the ne uroscien ce poin t o f view , althoug h much infor mation has been gathered , we are only star ting to grasp how language may be organiz ed in the human brain, and

each findi ng rais es new questio ns. Questions that also hold consider able clinical interest include the ne ural co rrelates of language learnin g and re-l earning in children and adu lts and the ov erall de velopm ent of language function from childhood through adolesc ence to adulthood. At present , the di erent componen ts of language functio n are typica lly probed using isol ated sti muli that allow good con trol of the task and stimulus pro perties but obv iously creat e rather unnatural expe rimental situa tions. W hile this is a well- grounded and necessa ry app roach, the brain correlates of language process ing may appear quite di erent in natural performan ce which the brain is tuned for. Mo ving towards increasingl y realist ic experi menta l d e signs wi ll be a intrigu- ing challen ge not only con ceptual ly but also methodol ogical- ly as it will requ ire new approac hes for da ta an alysis. Tool s for extracti ng netw orks of brain areas with correlated tim e courses of acti vation, instead of isolated acti ve areas, have been intro duced and succ essfully ap plied to analys is of the motor syst em, also in the dom ain of clinical resear ch (Gross et al., 2 001, 2002; Sch nitzler an d Gross , 2005 ). Thes e meth- ods are currently being develop ed for use with co gnitive tasks (Kujal a et al., 2006; Sa lmelin an d Kujal a, 2006 ) whi ch may, eventual ly, lead to clinical applications as well. The present revie w was focu sed on the contribu tion and usabilit y o f MEG in basic research and clini cal evaluat ion of language functi on. Research in this domain is also actively done using EEG (for recent reviews see, e.g. , Fried- erici, 2005; Hagoort , 2003; Kutas and Fede rmeier, 2000 ), fMRI and PET (for recent reviews see, e.g. , De´ monet et al., 2005; Job ard et al., 2003; Matth ews et al., 2003; Price and Crinion, 2005; Vign eau et al., 2006 ) and transcrania l magnetic stimu lation (TMS ) (e.g ., Dra¨ ger et al., 20 04 ). The specific importance of the MEG method is in the way it merge s goo d s p atial accu racy with accurat e real- time tracking of cortical acti vity, reveal ing both the sequ ential structure of ne ural a ctivatio n and the frequent ly obs erved overlap between time co urses of acti vation in dist inct brain areas. Com bined spatiotem poral informat ion is a valuabl e asset in function al ne uroimaging and , i n parti cular, when

ARTICLE IN PRESS

R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx

15

identi fying the neu ral correlates of function al disorde rs. For exampl e, as discus sed above for dyslex ia, it is impor- tant to know that the ne ural abn ormalit y i n reading appea rs as early as 150 ms afte r stimulus onset, at the poi nt when the brain first begins to treat visual sti muli as spec ific entities , such as letter- strings or faces. Location- based analys is using fMR I/PET poin ts to abnormal ities in the same general brain area in dyslex ia (Paul esu et al., 2001 ), proba bly in word- form analys is ( Cohen et al., 2000 ) – but based on locat ion onl y the exact functional role of this activa tion, or the significan ce of its absence in dy slexia, is dicul t to resol ve ( Price and Devlin, 200 3 ). A more detai led compari son of M E G and fMRI/P ET data reveal s small but intrigui ng di erence s i n anatom y and function of the left occipi totem poral acti vation in read ing, suggesting that MEG detects the onset of letter- string-s pecific analys is which is not detected in or does not dominat e the hemod y- namic signal . fMR I/PET, on the other han d, may detect subsequ ent acti vation along the ventral stream wher e ne u - rons woul d b e increa singl y sensitiv e t o the word-li keness of the letter- strings but their acti vation woul d b e less strong ly synchron ized or less rigor ously time-l ocked to stimu lus present ation and might thus go unde tected in M E G (or EEG ) (Corne lissen et al., 2003b ; Sa lmelin and Hele nius, 2004 ). As regards the sub sequent process ing stage s i n read- ing, namely, semant ic an d phonolo gical analys is, the agree- ment between MEG findi ngs and hemod ynami c measur es (see Jobard et al., 2003 for a meta -analysis of 3 5 fMRI/ PET studies) is not impr essive. It is current ly not known wheth er the amount of activit y (fMR I, PET) or its syn- chron icity (MEG, EEG) is more relev ant to human beh av- iour. Clea rly, for a comp rehensive pictur e o f the ne ural repres entation of language functi on it will be essential to unde rstand and capit alize on the sim ilarities an d di erenc es betwe en he modynami c a n d ne urophysi ologic al measur es.

Ackno wledgem ents

Financi al inte rests: This work was sup ported by the Cen- tre of Excell ence Program mes 200 0–2005 an d 2006–201 1 of the Academ y o f Finland, the Sigri d Juseliu s Found ation, and the James S. McDonn ell Found ation 21st Centu ry Re- search Award.

Reference s

Biermann-Ruben K, Salmelin R, Schnitzler A. Right rolandic activation during speech perception in stutterers: a MEG study. Neuroimage

2005;25:793–801.

Bloodstein O. A Handbook o n Stuttering. San Diego, CA: Singular Publishing Group Inc.; 1995. Bonte M, Parviainen T , Hyto¨ nen K, Salmelin R. Time course of top-down and bottom-up influences on syllable processing in the auditory cortex. Cereb Cortex 2006;16:115–23. Bowyer SM, M oran JE, Weiland BJ, Mason KM, Greenwald ML, Smith BJ, et al. Language laterality determined by MEG mapping with MR- FOCUSS. Epilepsy Behav 2005;6:235–41. Bradley L, Bryant PE. Categorizing sounds and learning to read – a causal connection. Nature 1983;301:419–21.

Breier JI, Castillo EM, Boake C, Billingsley R, Maher L, Francisco G , et al. Spatiotemporal patterns of language-specific brain activity in patients with chronic aphasia after stroke using magnetoencephalog- raphy. Neuroimage 2004;23:1308–16. Breier JI, Castillo EM, Simos PG, Billingsley-Marshall RL, Pataraia E, Sarkari S, et al. A typical language representation in patients with chronic seizure disorder and achievement deficits with magnetoen- cephalography. Epilepsia 2005;46:540–8. Breier JI, Simos PG, Fletcher JM, Castillo EM, Zhang W, Papanicolaou AC. Abnormal activation of temporoparietal language areas during phonetic analysis in children with dyslexia. Neuropsychology

2003;17:610–21.

Breier JI, Simos PG, Wheless JW, Constantinou JE, Baumgartner JE, Venkataraman V, et al. Language dominance in children as deter- mined by magnetic source imaging and the intracarotid amobarbital procedure: a comparison. J Child Neurol 2001;16:124–30. Breier JI, Simos PG, Zouridakis G, Papanicolaou AC. Lateralization o f cerebral activation in auditory verbal and non-verbal memory tasks using magnetoencephalography. Brain Topogr 1999a;12:89–97. Breier JI, Simos PG, Zouridakis G, Papanicolaou AC. Lateralization o f activity associated with language function using magnetoencephalog- raphy: a reliability study. J Clin Neurophysiol 2000;17:503–10. Breier JI, Simos PG, Zouridakis G, Wheless JW, W illmore LJ, Constan- tinou JE, et al. Language dominance determined by magnetic source imaging: a comparison with the Wada procedure. Neurology

1999b;53:938–45.

Butz M, Gross J, Timmermann L , M oll M, Freund HJ, Witte OW, et al. Perilesional pathological oscillatory activity in the magnetoencephalo- gram of patients with cortical brain lesions. Neurosci Lett

2004;355:93–6.

Caramazza A, Hillis A. Lexical organization o f nouns and verbs in the brain. Nature 1991;349:788–90. Cohen L, Dehaene S , Naccache L, Lehericy S, Dehaene-Lambertz G, Hena MA, e t al. The visual word form area – spatial and temporal characterization of an initial stage of reading in normal subjects and posterior split-brain patients. Brain 2000;123:291–307. Coltheart M, Curtis B, Atkins P, Haller M. Models of reading aloud:

dual-route and parallel-distributed-processing approaches. Psychol Rev 1993;100:589–608. Connolly JF, Phillips NA. Event-related potential components reflect

phonological and semantic processing of the terminal word of spoken sentences. J Cogn Neurosci 1994;6:256–66. Cornelissen K, Laine M, Tarkiainen A , J a¨ rvensivu T, Martin N, Salmelin

R. Adult brain plasticity elicited by anomia treatment. J Cogn

Neurosci 2003a;15:444–61. Cornelissen P, Tarkiainen A, Helenius P, Salmelin R. Cortical e ects of

shifting letter-position in letter-strings of varying length. J Cogn Neurosci 2003b;15:731–46. Dale AM, Liu AK, Fischl BR, Buckner RL, Belliveau JW, Lewine JD, et al. Dynamic statistical parametric mapping: combining fMRI and MEG for high-resolution imaging of cortical activity. Neuron 2000;26:55–67. De´ monet JF, Thierry G, Cardebat D. Renewal of the neurophysiology of language: functional neuroimaging. Physiol Rev 2005;85:49–95. de Jongh A, Baayen JC, de Munck JC, Heethaar RM, Vandertop WP, Stam CJ. The influence of brain tumor treatment on pathological delta activity in MEG. Neuroimage 2003;20:2291–301. Dra¨ ger B, Breitenstein C, Helmke U, Kamping S, Knecht S. Specific and

nonspecific e ects of transcranial magnetic stimulation o n picture- word verification. Eur J Neurosci 2004;20:1681–7. Eulitz C, Diesch E, Pantev C, Hampson S, Elbert T. Magnetic and electric brain activity evoked by the processing of tone and vowel stimuli. J Neurosci 1995;15:2748–55. Fairbanks G. Systematic research in experimental phonetics: 1. A theory of the speech mechanism as a servosystem. J Speech Hear Disord

1954;19:133–9.

Fisher AE, Barnes GR, Hillebrand A, Holliday IE, Witton C, Richards

IL. Abnormality of mismatch negativity in response to tone omission

in dyslexic adults. Brain Res 2006;1077:90–8.

ARTICLE IN PRESS

16

R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx

Fiez JA, Petersen SE. Neuroimaging studies of word reading. Proc Natl Acad Sci USA 1998;95:914–21. Friederici AD. Neurophysiological markers of early language acquisition:

from syllables to sentences. Trends Cogn Sci 2005;9:481–8. Gootjes L, Raiji T, Salmelin R, Hari R. Left-hemisphere dominance for processing of vowels: a whole-scalp neuromagnetic study. Neuroreport

1999;10:2987–91.

Gross J, Kujala J, Ha¨ ma¨ la¨ inen M, Timmermann L , Schnitzler A, Salmelin R. Dynamic imaging of coherent sources: studying neural interactions in the human brain. Proc Natl Acad Sci USA 2001;98:694–9. Gross J, Timmermann J, Kujala J, Dirks M, Schmitz F, Salmelin R, et al. The neural basis of intermittent motor control in humans. Proc Natl Acad Sci USA 2002;99:2299–302. Hagoort P. How the brain solves the binding problem for language: a neurocomputational model of syntactic processing. Neuroimage 2003;20(Suppl 1):S18–29. Halgren E, Baudena P, Heit G, Clarke M, Marinkovic K. Spatio-temporal stages in face and word processing. 1. Depth-recorded potentials in the human occipital, temporal and parietal lobes. J Physiol (Paris)

1994;88:1–50.

Halgren E, Dhond RP, Christensen N, Van Petten C, Marinkovic K, Lewine JD, et al. N400-like magnetoencephalography responses modulated b y semantic context, word frequency, and lexical class in sentences. Neuroimage 2002;17:1101–16. Hari R. The neuromagnetic method in the study of the human auditory cortex. In: Grandori F, Hoke M, Romani GL, editors. Auditory Evoked Magnetic Fields and Electric Potentials, vol. 6. Basel: S. Karger; 1990. p. 222–82. Hari R, Salmelin R. Human cortical oscillations: a neuromagnetic view through the skull. Trends Neurosci 1997;20:44–9. Heim S, Eulitz C, Elbert T. Altered hemispheric asymmetry of auditory N100m in adults with developmental dyslexia. Neuroreport 2003a;14:501–4. Heim S, Eulitz C, Elbert T. Altered hemispheric asymmetry of auditory P100m in dyslexia. Eur J Neurosci 2003b;17:1715–22. Helenius P, Salmelin R, Richardson U, Leinonen S, Lyytinen H. Abnormal auditory cortical activation in dyslexia 100 ms after speech onset. J Cogn Neurosci 2002a;14:603–17. Helenius P, Salmelin R, Service E, Connolly JF. Distinct time courses of word and sentence comprehension in the left temporal cortex. Brain

1998;121:1133–42.

Helenius P, Salmelin R, Service E, Connolly JF. Semantic cortical activation in dyslexic readers. J Cogn Neurosci 1999a;11:535–50. Helenius P, Salmelin R, Service E, Connolly JF, Leinonen S, Lyytinen H. Cortical activation during spoken-word segmentation i n nonreading- impaired and dyslexic adults. J Neurosci 2002b;22:2936–44. Helenius P, Tarkiainen A, Cornelissen P, Hansen PC, Salmelin R. Dissociation o f normal feature analysis and deficient processing of letter-strings in dyslexic adults. Cereb Cortex 1999b;9:476–83. Hickok G, Poeppel D. Dorsal and ventral streams: a framework for understanding aspects of the functional anatomy of language. Cogni- tion 2004;92:67–99. Hirata M, Kato A, Taniguchi M, Saitoh Y, Ninomiya H, Ihara A, et al. Determination of language dominance with synthetic aperture magne- tometry: comparison with the Wada test. Neuroimage 2004;23:46–53. Ha¨ ma¨ la¨ inen M, Hari R, Ilmoniemi RJ, Knuutila J, Lounasmaa OV. Magnetoencephalography – theory, instrumentation, and applications to noninvasive studies of the working human brain. Rev Mod Phys

1993;65:413–97.

Indefrey P, Levelt WJ. The spatial and temporal signatures of word production components. Cognition 2004;92:101–44. Jobard G, Crivello F, Tzourio-Mazoyer N. Evaluation of the dual route theory of reading: a metanalysis of 35 neuroimaging studies. Neuro- image 2003;20:693–712. Kamada K, Sawamura Y, Takeuchi F, Houkin K, Kawaguchi H, Iwasaki Y, et al. Gradual recovery from dyslexia and related serial magneto- encephalographic changes in the lexicosemantic centers after resection of a mesial temporal astrocytoma. Case report. J Neurosurg

2004;100:1101–6.

Kirveskari E , Salmelin R, Hari R. Neuromagnetic responses to vowels vs. tones reveal hemispheric lateralization. Clin Neurophysiol

2006;117:643–8.

Kober H, Moller M, Nimsky C, Vieth J, Fahlbusch R, Ganslandt O. New approach to localize speech relevant brain areas and hemispheric dominance using spatially filtered magnetoencephalography. Hum Brain Mapp 2001;14:236–50. Kujala A, Alho K, Service E, Ilmoniemi RJ, Connolly JF. Activation in the anterior left auditory cortex associated with phonological analysis of speech input: localization of the phonological mismatch negativity response with MEG. Brain Res Cogn Brain Res

2004;21:106–13.

Kujala J, Pammer K , Cornelissen P, Roebroeck A, Formisano E, Salmelin R. Phase coupling in a cerebro-cerebellar network at 8–13 Hz during reading. Cereb Cortex 2006. Epub doi:10.1093/cercor/bhl059. Kuriki S, Murase M. Neuromagnetic study of the auditory responses in right and left hemispheres of the human brain evoked by pure tones and speech sounds. Exp Brain Res 1989;77:127–34. Kutas M, Hillyard SA. Reading senseless sentences: brain potentials reflect semantic incongruity. Science 1980;207:203–5. Kutas M, Federmeier KD. Electrophysiology reveals semantic memory use in language comprehension. Trends Cogn Sci

2000;4:463–70.

Laine M, Salmelin R, Helenius P, Marttila R. Brain activation during reading in deep dyslexia: an MEG study. J Cogn Neurosci

2000;12:622–34.

Levelt WJ, Praamstra P, Meyer AS, Helenius P, Salmelin R. An MEG study of picture naming. J Cogn Neurosci 1998;10:553–67. Levelt WJM, Roelofs A, Meyer AS. A theory of lexical access in speech production. Behav Brain Sci 1999;22:1–38. Maess B, Friederici AD, Damian M , Meyer AS, Levelt WJ. Semantic category interference in overt picture naming: sharpen- ing current density localization by PCA. J Cogn Neurosci

2002;14:455–62.

Maestu F, Ortiz T, Fernandez A , Amo C, Martin P, Fernandez S , e t al. Spanish language mapping using MEG: a validation study. Neuroim- age 2002;17:1579–86. Marinkovic K, Dhond RP, Dale AM, Glessner M, Carr V, Halgren E. Spatiotemporal dynamics of modality specific and supramodel word processing. Neuron 2003;38:487–97. Matthews P M, Adcock J, Chen Y, Fu S, Devlin JT, Rushworth MF, e t al. Towards understanding language organisation i n the brain using fMRI. Hum Brain Mapp 2003;18:239–47. Meinzer M, Elbert T, Wienbruch C, Djundja D, Barthel G, Rockstroh B. Intensive language training enhances brain plasticity in chronic aphasia. BMC Biol 2004;2:20. Ma¨ kela¨ JP, Kirveskari E, Seppa M, Ha¨ ma¨ la¨ inen M, Forss N, Avikainen S, et al. Three-dimensional integration of brain anatomy and function to facilitate intraoperative navigation around the sensorimotor strip. Hum Brain Mapp 2001;12:180–92. Nagarajan S, Mahncke H, Salz T, Tallal P, Roberts T, Merzenich MM. Cortical auditory signal processing in poor readers. Proc Natl Acad Sci USA 1999;96:6483–8. Nobre AC, Allison T, McCarthy G. Word recognition i n the human inferior temporal lobe. Nature 1994;372:260–3. Nobre AC, McCarthy G. Language-related field potentials in the anterior- medial temporal lobe: II. E ects of word type and semantic priming. J Neurosci 1995;15:1090–8. Na¨ a¨ ta¨ nen R, Lehtokoski A, Lennes M, Cheour M, Huotilainen M, Iivonen A, et al. Language-specific phoneme representations revealed by electric and magnetic brain responses. Nature

1997;385:432–4.

Na¨ a¨ ta¨ nen R, Jacobsen T , Winkler I. Memory-based or a erent processes in mismatch negativity (MMN): a review of the evidence. Psycho- physiol 2005;42:25–32. Obleser J, Elbert T, Eulitz C. Attentional influences on functional mapping of speech sounds in human auditory cortex. BMC Neurosci

2004;5:24.

ARTICLE IN PRESS

R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx

17

Pammer K, Hansen PC, Kringelbach ML, Holliday I, Barnes G, Hillebrand A , e t al. Visual word recognition: the first half second. Neuroimage 2004;22:1819–25. Pantev C, Ross B, Berg P, Elbert T, Rockstroh B. Study of the human auditory cortices using a whole-head magnetometer: left vs. right hemisphere and ipsilateral vs. contralateral stimulation. Audiol Neu- rootol 1998;3:183–90. Papanicolaou AC, Simos PG, Castillo EM, Breier JI, Sarkari S, Pataraia E, et al. Magnetoencephalography: a noninvasive alternative to the Wada procedure. J Neurosurg 2004;100:867–76. Parviainen T, Helenius P , Poskiparta E, Niemi P, Salmelin R. Cortical sequence of word perception in beginning readers. J Neurosci

2006;26:6052–61.

Parviainen T, Helenius P, Salmelin R. Cortical di erentiation of speech and nonspeech sounds at 100 ms: implications for dyslexia. Cereb Cortex 2005;15:1054–63. Pataraia E, Billingsley-Marshall RL, Castillo EM, Breier JI, Simos PG, Sarkari S, et al. Organization o f receptive language-specific cortex before and after left temporal lobectomy. Neurology 2005;64:481–7. Pataraia E, Simos PG, Castillo EM, Billingsley-Marshall RL, McGregor AL, Breier JI, et al. Reorganization o f language-specific cortex in patients with lesions or mesial temporal epilepsy. Neurology

2004;63:1825–32.

Paul I, Bott C, Heim S, Eulitz C, Elbert T. Reduced hemispheric asymmetry of the auditory N260m in dyslexia. Neuropsychologia

2006;44:785–94.

Paulesu E, Demonet JF, Fazio F, McCrory E, Chanoine V, Brunswick N, et al. Dyslexia: cultural diversity and biological unity. Science

2001;291:2165–7.

Phillips C, Pellathy T, Marantz A, Yellin E, Wexler K, Poeppel D, et al. Auditory cortex accesses phonological categories: an MEG mismatch study. J Cogn Neurosci 2000;12:1038–55. Plaut D, McClelland J, Seidenberg M, Patterson K. Understanding normal and impaired word reading: computational principles in quasi- regular domains. Psychol Rev 1996;103:56–115. Poeppel D, Yellin E, Phillips C, Roberts TPL, Rowley HA, Wexler K, et al. Task-induced asymmetry of the auditory evoked M100 neuro- magnetic field elicited by speech sounds. Brain Research

1996;4:231–42.

Price CJ, Crinion J. The latest on functional imaging studies of aphasic stroke. Curr Opin Neurol 2005;18:429–34. Price CJ, Devlin JT. The myth of the visual word form area. Neuroimage

2003;19:473–81.

Pylkka¨ nen L, Llinas R, Murphy GL. The representation of polysemy:

MEG evidence. J Cogn Neurosci 2006;18:97–109. Pylkka¨ nen L, Marantz A. Tracking the time course of word recognition with MEG. Trends Cogn Sci 2003;7:187–9. Pylkka¨ nen L, Stringfellow A, Marantz A. Neuromagnetic evidence for the timing of lexical activation: a n M EG component sensitive to phono- tactic probability but not to neighborhood density. Brain Lang

2002;81:666–78.

Renvall H, Hari R. Auditory cortical responses to speech-like stimuli in dyslexic adults. J Cogn Neurosci 2002;14:757–68. Renvall H, Hari R. Diminished auditory mismatch fields in dyslexic adults. Ann Neurol 2003;53:551–7. Saarinen T, Laaksonen H, Parviainen T, Salmelin R. Motor cortex dynamics in visuomotor production of speech and non-speech mouth movements. Cereb Cortex 2006;16:212–22. Salmelin R, Hari R. Spatiotemporal characteristics of sensorimotor MEG rhythms related to thumb movement. Neuroscience

1994;60:537–50.

Salmelin R, Hari R, Lounasmaa O , Sams M. Dynamics o f brain activation during picture naming. Nature 1994;368:463–5. Salmelin R, Helenius P. Functional neuroanatomy of impaired reading in dyslexia. Sci Studies Reading 2004;8:257–72. Salmelin R, Helenius P , Service E. Neurophysiology of fluent and impaired reading: a magnetoencephalographic approach. J Clin Neurophysiol 2000a;17:163–74.

Salmelin R, Ha¨ ma¨ la¨ inen M, Kajola M, Hari R. Functional segregation of movement-related rhythmic activity in the human brain. Neuroimage

1995;2:237–43.

Salmelin R, Kujala J. Neural representation of language: activation vs. long-range connectivity. Trends Cogn Sci 2006. Salmelin R, Sams M. Motor cortex involvement during verbal versus non-verbal lip and tongue movements. Hum Brain Mapp

2002;16:81–91.

Salmelin R, Schnitzler A, Parkkonen L, Biermann K , H elenius P , Kiviniemi K, e t al. Native language, gender, and functional organi- zation of the auditory cortex. Proc Natl Acad Sci USA

1999;96:10460–5.

Salmelin R, Schnitzler A, Schmitz F, Freund HJ. Single word reading in developmental stutterers and fluent speakers. Brain

2000b;123:1184–202.

Salmelin R, Schnitzler A, Schmitz F, Ja¨ ncke L, Witte O, Freund H-J. Functional organization o f the auditory cortex is di erent in stutterers and fluent speakers. Neuroreport 1998;9:2225–9. Salmelin R, Service E, Kiesila¨ P, Uutela K, Salonen O. Impaired visual word processing in dyslexia revealed with magnetoencephalography. Ann Neurol 1996;40:157–62. Schnitzler A, Gross J. Normal and pathological oscillatory communica- tion in the brain. Nat Rev Neurosci 2005;6:285–96. Seidenberg MS, M cClelland JL. A distributed, developmental model of word recognition and naming. Psychol Rev 1989;96:523–68. Shaywitz SE, Shaywitz BA, Pugh KR, Fulbright RK, Constable RT, Mencl WE, et al. Functional disruption in the organization o f the brain for reading in dyslexia. Proc Natl Acad Sci USA

1998;95:2636–41.

Shtyrov Y, Kujala T, Lyytinen H, Ilmoniemi RJ, Na¨ a¨ ta¨ nen R. Auditory cortex evoked magnetic fields and lateralization of speech processing. Neuroreport 2000;11:2893–6. Simos PG, Basile LF, Papanicolaou AC. Source localization of the N400 response in a sentence-reading paradigm using evoked magnetic fields and magnetic resonance imaging. Brain Res 1997;762:29–39. Simos PG, Breier JI, Fletcher JM, Bergman E, Papanicolaou AC. Cerebral mechanisms involved in word reading in dyslexic children: a magnetic source imaging approach. Cereb Cortex 2000a;10:809–16. Simos PG, Breier JI, Fletcher JM, Foorman BR, Bergman E, Fishbeck K, et al. Brain activation profiles in dyslexic children during non-word reading: a magnetic source imaging study. Neurosci Lett

2000b;290:61–5.

Simos PG, Breier JI, Zouridakis G, Papanicolaou AC. Assessment of functional cerebral laterality for language using magnetoencephalog- raphy. J Clin Neurophysiol 1998;15:364–72. Simos PG, Fletcher JM, Bergman E, Breier JI, Foorman BR, Castillo EM, et al. Dyslexia-specific brain activation profile becomes normal following successful remedial training. Neurology

2002a;58:1203–13.

Simos PG, Fletcher JM, Foorman BR, Francis DJ, Castillo EM, Davis RN, et al. Brain activation profiles during the early stages of reading acquisition. J Child Neurol 2002b;17:159–63. Simos PG, Papanicolaou AC, Breier JI, Wheless JW, Constantinou JE, Gormley WB, et al. Localization of language-specific cortex by using magnetic source imaging and electrical stimulation m apping. J Neurosurg 1999;91:787–96. Sommer M, Koch MA, Paulus W, Weiller C, Bu¨ chel C. A disconnection of speech-relevant brain areas in developmental stuttering. Lancet

2002;360:380–3.

Szymanski MD, Rowley HA, Roberts TP. A hemispherically asymmet- rical MEG response to vowels. Neuroreport 1999;10:2481–6. So¨ ro¨ s P , Cornelissen K, Laine M, Salmelin R. Naming actions and objects: cortical dynamics in healthy adults and in an anomic patient with a dissociation i n action/object naming. Neuroimage

2003;19:1787–801.

Tarkiainen A, Cornelissen PL, Salmelin R. Dynamics of visual feature analysis and object-level processing in face versus letter-string percep- tion. Brain 2002;125:1125–36.

ARTICLE IN PRESS

18

R. Salmelin / Clinical Neurophysiology xxx (2006) xxx–xxx

Tarkiainen A, Helenius P, Hansen PC, Cornelissen PL, Salmelin R. Dynamics o f letter string perception in the human occipitotemporal cortex. Brain 1999;122:2119–32. Tarkiainen A, Helenius P, Salmelin R. Category-specific occipitotemporal activation during face perception in dyslexic individuals: an MEG study. Neuroimage 2003;19:1194–204. Tiitinen H, Sivonen P, Alku P, Virtanen J, Na¨ a¨ ta¨ nen R. Electromagnetic recordings reveal latency di erences in speech and tone processing in humans. Brain Res Cogn Brain Res 1999;8:355–63. Trauzettel-Klosinski S, Durrwachter U, Klosinski G , Braun C. Cortical activation during word reading and picture naming in dyslexic and non-reading-impaired children. Clin Neurophysiol 2006;117:1085–97.

M, Somersalo E. Visualization of magnetoen-

cephalographic data using minimum current estimates. Neuroimage

Uutela K, Ha¨ ma¨ la¨ inen

1999;10:173–80.

Vigneau M, Beaucousin V, Herve´ PY, Du au H, Crivello F, Houde´ O, Mazoyer B, Tzourio-Mazoyer N. Meta-analyzing left processing. Neuroimage 2006;30:1414–32.

Vihla M, Lounasmaa OV, Salmelin R. Cortical processing of change detection: dissociation between natural vowels and two-frequency complex tones. Proc Natl Acad Sci USA 2000;97:10590–4. Vihla M, Salmelin R. Hemispheric balance in processing attended and non-attended vowels and complex tones. Cogn Brain Res

2003;16:167–73.

Walla P, Mayer D, Deecke L, Thurner S. The lack of focused anticipation of verbal information in stutterers: a magnetoencephalographic study. Neuroimage 2004;22:1321–7. Wilson TW, Leuthold AC, Lewis SM, Georgopoulos AP, Pardo PJ. The time and space of lexicality: a neuromagnetic view. Exp Brain Res

2005;162:1–13.

Wydell TN, Vuorinen T, Helenius P, Salmelin R. Neural correlates of letter-string length and lexicality during reading in a regular orthog- raphy. J Cogn Neurosci 2003;15:1052–62. Zouridakis G, Simos PG, Breier JI, Papanicolaou AC. Functional hemispheric asymmetry assessment in a visual language task using MEG. Brain Topogr 1998;11:57–65.