A

iawtnlwloTd©

K

1

rtssftstahfe

8

0d

Behavioural Brain Research 175 (2006) 296–304

Research report

Language lateralization in phonological, semantic andorthographic tasks: A slow evoked potential study

Chiara Spironelli, Alessandro Angrilli ∗Department of General Psychology, University of Padova, Padova, Italy

Received 29 May 2006; received in revised form 23 August 2006; accepted 28 August 2006Available online 12 October 2006

bstract

Most of literature on language has shown how different word-classes activate distinct neural networks within linguistic cortical areas. The presentnvestigation aimed to demonstrate that, by means of slow evoked potentials and using the same set of words in different tasks, it is possible toctivate cortical networks that are spatially and temporally distinguished. Twenty healthy subjects had to evaluate, in a word pair matching session,hether two words rhymed (phonological task), were semantically related (semantic task) or were written in the same letter case (orthographic

ask). Slow wave amplitude was computed in three relevant time windows: the last 0.5 s of first word presentation (W1), the initial contingentegative variation (iCNV) and the terminal CNV (tCNV). During W1 and iCNV intervals, both the orthographic and the phonological tasks were leftateralized. Furthermore, the phonological task was more lateralized than the orthographic because of a greater inhibition of the right hemisphere,hereas the orthographic task was characterized by a greater bilateral posterior activation. During the tCNV, only the phonological task remained

eft lateralized while orthographic and semantic were bilaterally distributed. Although the use of the same set of words tends to activate widelyverlapped networks, in the present research task manipulation was effective in demonstrating task dependent differences in brain lateralization.hus, the present paradigm and the adopted tasks are especially suited for studying deficit and recovery of language in patients affected by linguisticisorders such as developmental dyslexia and aphasia.

2006 Elsevier B.V. All rights reserved.

oding

dopiificcAoat

eywords: Language lateralization; Slow cortical potentials; Phonological enc

. Introduction

The functional lateralization of the human brain seems toepresent a prerequisite for the full realization of the linguis-ic potential [28,33,37]. Functional neuroimaging techniques,uch as fMRI and PET, have clearly demonstrated that the mostteady laterality effects are reliably located over left inferiorrontal areas [9,12,19–21,27,36,42,44,50,56,64,67–69]. Never-heless, although functional techniques provide a very goodpatial resolution (fMRI and PET) and a proved measure ofhe laterality index (fTCD and PET), the time course of corticalctivation cannot be studied with precision, a problem which

ighlights the need of complementary non-invasive methodsor the full assessment of language lateralization. It is evident,specially in the linguistic field, the need for distinguishing the

∗ Corresponding author at: Department of General Psychology, Via Venezia, 35131 Padova, Italy. Tel.: +39 049 8276692; fax: +39 049 8276600.

E-mail address: alessandro.angrilli@unipd.it (A. Angrilli).

dEsett(s

166-4328/$ – see front matter © 2006 Elsevier B.V. All rights reserved.oi:10.1016/j.bbr.2006.08.038

; Semantic processing; ERPs; EEG

ifferent processes that occur across time also within few tensf milliseconds, and that usually in a high spatial resolutionicture of brain metabolism appear to be largely overlappedn the temporal dimension (see [43] for a critical review). Its possible, for instance, that lateralization of language is con-ned, depending on the task and the selected stimuli, to earlyompared with late phases of word processing, an issue thatannot be addressed with methods based on brain metabolism.lthough EEG does not provide a comparable spatial resolutionf brain activity, a reliable discrimination of electrical activityt level of gross cortical regions such as anterior versus pos-erior and left versus right hemisphere can be obtained. Thus,epending on the scientific question it is possible to obtain fromEG reliable and relevant information on lateralization by usingpecial care in the selection of methods and paradigms. Sev-ral studies have been carried out using ERPs to analyze the

ime course and the distribution of the electrical activity overhe scalp of a number of linguistic functions, both in adultse.g. [2,26,34,46,48,54,55]) and in functionally recovered apha-ic patients in whom a substantially affected lateralization was

ral B

eotuilaawraswaticagmtdtlfssapwtewccl

tatpbrctGtctstbttogw

adtlgthpwtrlelr

2

2

rAEfm

2

cwhrw(aaNrt

2

fwifbpcwpd[u[j

C. Spironelli, A. Angrilli / Behaviou

xpected (e.g. [3,5,6,22,23,25,31,63,65]). However, only a fewf the quoted investigations have used an ERP paradigm to assesshe lateralization in specific linguistic tasks. For example, bysing a prime-target matching paradigm, Rugg [54,55] has foundn adults a left lateralized slow ERP negativity during a phono-ogical judgement task, whereas Altenmuller et al. [3], as wells Thomas et al. [65] have recorded, during a synonym gener-tion task in aphasic patients, slow negative potentials whichere greater over the left than the right frontocentral cortical

egions. Other studies have shown how different word classesre able to activate distinct cortical regions within the left hemi-phere that are related with specific word functions. For instance,ords related to action activate, in addition to linguistic areas,

lso the motor cortex, whereas visual-related words activateemporal regions where the visual representation of the words stored; furthermore, different networks are involved in theortical elaboration of function versus content words [47], andlso mass and count nouns are differently distributed across lin-uistic regions [14]. However, in task-focused experiments, theentioned word categories cannot be easily balanced for seman-

ic relevance [57], frequency, length, phonology, etc., making itifficult to disentangle their parametric class-specific proper-ies, especially in view of their use in clinical populations withinguistic deficits that include also class-specific deficits. As aurther problem, it is necessary to emphasize that, in most EEGtudies, electrical activity has been recorded with only a fewelected electrodes referred to linked mastoids, thus determiningn important bias. Indeed, it is currently well known that the tem-oral cortex underlying mastoid sites is an active brain structurehich is strongly involved in most linguistic processes. Fur-

hermore, the most common eye movement correction method,.g. the regressive analysis [29,59], tends to remove, togetherith ocular artifacts, also the electrical activity of anterior corti-

al regions, thus providing an underestimation of activity fromortical areas (e.g. frontal and temporal) critically involved inanguage.

The linguistic paradigm of the present research is based onhe use of the same sample of concrete words in different tasks,methodological strategy that should strongly reduce most of

he limits listed for other approaches. We have validated an ERParadigm to test linguistic lateralization in different tasks [4]y using both a language-free scalp reference, that is averageeference, and an eye artifact modelling method to optimize theorrection of all recorded channels [10,11], including frontal andemporal sites. In that study, we have found in both Italian anderman native speakers significant marked left lateralization of

he evoked potentials measured over the frontal and temporalortices during the rhyming task, but not in lexical–semanticasks [4]. With respect to our past experiment, in the presenttudy we introduced three main innovations. First, we comparedhree different linguistic tasks, by using the orthographic as moreasic control task that better contrast the activity induced by bothhe semantic and the phonological tasks. Second, we analyzed

hree specific time intervals, corresponding to different phasesf word elaboration, to follow the temporal development of lin-uistic elaboration. Finally, we selected a group of aged adultsith respect to the young adults of the past sample with the

fiItiw

rain Research 175 (2006) 296–304 297

im to verify the generalization of lateralization effect acrossifferent ages. In line with prior evidence, the current investiga-ion was carried out to further validate and generalize our CNVanguage-related paradigm in measuring lateralization of lin-uistic processes. According to prior findings [4], we expectedhe phonological task to elicit a greater activation of the leftemisphere and a relative inhibition of the right hemisphere,articularly at level of frontal cortex, whereas the semantic taskas expected (starting from EEG rather than from fMRI litera-

ure) to induce a bilateral more distributed cortical activity. Witheference to the orthographic control task, we foresaw an initialeft-lateralized cortical pattern produced by the automatic wordlaboration (even though unnecessary to perform this task), fol-owed by a shift of strategy related to the main non-linguisticequirements of the task.

. Materials and methods

.1. Participants

Twenty native Italian subjects (seven females, mean age 59.10 ± 7.11 years,ange 43–69 years) gave their written consent to participate in the experiment.ll subjects were fully right-handed, in average above 90% according to thedinburgh Handedness Inventory [38]. None of the subjects had been treated

or any neurological or psychiatric disorder, nor was under medication. Experi-ental procedures were approved by the local Ethics Committee.

.2. Apparatus and physiological recordings

In order to improve our past paradigm, we increased both the number ofortical sites recorded, and the temporal resolution. Thus, electroencephalogramas continuously collected in dc mode, by following the main requirements forigh quality dc recordings [7], with a low-pass filter set to 30 Hz, samplingate of 500 Hz, and amplitude resolution of 0.168 �V/bin. Electrical signalsere measured by means of 38 tin electrodes, using two SynAmps amplifiers

NeuroScan Labs, Sterling, USA), 31 mounted on an elastic cap (ElectroCap)ccording to the International 10-20 system [39]; the other seven electrodes werepplied below each eye (Io1, Io2), on the two external canthii (F9, F10), on theasion (Nz) and on mastoids (M1, M2; see Fig. 1b). Cz was used as on-line

ecording reference for the EEG channels, then data were converted off-line tohe average reference.

.3. Stimuli, tasks and procedure

Stimuli consisted of bi- and tri-syllabic Italian content words with averagerequency, selected from a dictionary of 5000 written Italian words [16]. Wordsere visually presented in pairs on a computer screen one at a time, with an

nter-stimulus interval (ISI) of 2 s: the first word (W1) remained on the screenor 1 s, the second word (W2 or target) was presented until subjects respondedy pressing a keyboard button, in any case not longer than 5 s (Fig. 1a). Wordairs were used in three tasks administered in separated blocks. To avoid psy-hophysical and word-category systematic effects, the same sample of wordsas presented as W1 in all tasks, but in different randomized order. Indeed,revious cross-linguistic validation studies showed that the same words activateifferent neural networks according to the task in which subjects are involved4,24]. Thus, in the phonological encoding task, participants had to decide,pon W2-target presentation, whether the word pairs rhymed (e.g. brodo-chiodobroth-spike]) or not (e.g. neve-corda [snow-rope]). In the semantic task, sub-ects had to decide whether the target word W2 was semantically related to the

rst (e.g. brodo-minestra [broth-soup]) or not (e.g. neve-sveglia [snow-alarm]).n the orthographic control task, participants were asked to press a “yes” but-on if both words were written in the same types (either lower or upper cases,.e. BRODO-FRUTTA [BROTH-FRUIT]), or a “no” button if the two wordsere written in different types (lower versus upper cases or vice versa, i.e. neve-

298 C. Spironelli, A. Angrilli / Behavioural B

Fig. 1. Diagram showing (a) the trial structure with W1, the first word presentedfor 1 s, inter-stimulus interval (ISI), inter-trial interval (ITI), W2, the secondword which lasted until a button was pressed and not longer than 5 s; (b) the fourcPi

PmEwww

2

oacillc[sittap1inWriio[[

np

aararFPawgpaitaf

rst1lo

espw

3

3

(ttvrh

3

ersiaamsba

lusters of electrodes, AL, anterior left, AR, anterior right, PL, posterior left andR, posterior right which entered in statistics, and the central cluster (electrodes

n italic) which entered in the CNV analysis.

ALESTRA [snow-GYMNASIUM]). Subjects provided the motor response foratch/mismatch decision by pressing a button with left index or middle finger.very task included 80 trials/word-pairs, lasted for approximately 10 min andas preceded by a quick training session of 20 trails. In all tasks, 50% matchesere randomly interspersed with 50% mismatch trials. The order of the tasksas randomly assigned across subjects.

.4. Data analysis

In dc recordings, EEG traces can include very slow and random dc driftsf amplifiers, therefore in the early processing stage of ERP data, we correctednd removed dc shifts by means of a fourth order polynomial detrending pro-edure applied to the whole EEG file. Then the continuous signal was epochedn 13-s intervals (including 1 s before and 12 s after W1 onset), and a furtherinear detrend was performed on the segmented waveforms. A 100 ms base-ine preceding W1 was subtracted from the whole epoch, then raw data wereorrected for blinks and eye movement artifacts, according to Berg and Scherg10,11], by using BESA software (Brain Electrical Source Analysis, 5.1 ver-ion). Trials with residual artifacts were then visually inspected and rejectedf necessary. Trials corresponding to wrong behavioral responses were rejectedoo. Therefore, all accepted trials (67.5% on average) were averaged for eachask and for every subject. Three functional intervals have been chosen for datanalysis which correspond to specific phases of word elaboration: the 500 msreceding W1 offset, the first 1000 ms of the ISI after word offset, the last000 ms of the ISI. According to other studies which investigated slow ERPsn the W1–W2 paradigm [4,13,49,51,53], we separated the early CNV compo-ent, termed initial CNV (iCNV), corresponding to the 1–2 s interval following1 offset, from the late CNV component, termed terminal CNV (tCNV), cor-

esponding to the 2–3 s interval following W1 offset. Thus, we referred to theCNV as an index of cognitive operations closely related to the stimulus encod-

ng [52,53] in the verbal working memory [13,51] and to the tCNV as an indexf rehearsal of the first word W1 and preparation of the motor response/decision13,49,51]. Maps were obtained through spline interpolation methods41].

swbF

rain Research 175 (2006) 296–304

According to past literature on slow evoked potentials and CNV, the corticalegativity was considered an index of relative brain activation and the corticalositivity was indicative of relative brain inhibition [4,13,49].

Task-specific lateralization pattern of the CNV was evaluated by means ofnalyses of variance (ANOVAs) comparing, for each time interval, the aver-ge amplitude measured in four groups of electrodes, each representing aegion of interest. On the basis of prior experiment, the Global Field Powernalysis and the visual inspection of waveforms (Fig. 5), we selected fourepresentative electrodes for each quadrant: anterior left (AL: F3, F7, FT7,C3), anterior right (AR: F4, F8, FT8, FC4), posterior left (PL: P3, CP3,7, TP7) and posterior right (PR: P4, CP4, P8, TP8). The average electricalctivity of all electrodes was computed in every quadrant (Fig. 1b). Threeithin-subjects factors were entered in ANOVAs: task (three levels: ortho-raphic versus phonological versus semantic), region (two levels: anterior versusosterior) and laterality (two levels: left versus right hemisphere). We addedlso the between-subjects factor sex (two levels: males versus females) asn the literature there is limited evidence of a sex difference in lateraliza-ion of phonological processing [60]. Although we had no specific hypothesisnd expected no gender significant difference, we decided to control also thisactor.

In order to verify the effective CNV development into the 2 s ISI, we sepa-ately analyzed also the central electrodes in which this slow wave classicallyhows the greatest amplitude (that is in FCz, C3, Cz, C4 and CPz; see Fig. 1b). Tohis aim, we performed separated ANOVAs for the three time intervals (0.5–1,–2 and 2–3 s after W1 onset) by including the between-subjects sex factor (twoevels: males versus females) and the within-subjects task factor (three levels:rthographic versus phonological versus semantic).

Performance was examined by comparing mean response times (RTs) andrror rates, between tasks and groups (orthographic versus phonological versusemantic, and males versus females, respectively). Tukey HSD tests were com-uted for all post hoc comparisons (α = 0.05) and the Huynh–Feldt correctionas applied when necessary [35].

. Results

.1. Behavioral performance

Response times showed a significant main effect of taskF2,36 = 14.01, HF ε = 1.0, p < 0.001), exhibiting faster RTs tohe orthographic than to both the phonological (p < 0.01) andhe semantic tasks (p < 0.001), without sex differences (895 msersus 1053 and 1135 ms, respectively; see Fig. 2a). Error rateseflected these differences between tasks, with non-significantigher rates for the semantic task (see Fig. 2b).

.2. ERP regional asymmetry

In the first time interval (0.5–1 s), we found a significant mainffect only for laterality factor (F1,18 = 8.76, p < 0.01), whichevealed greater negativity in the left than in the right hemi-phere (−0.23 �V versus 0.39 �V, respectively). Two significantnteractions task by region (F2,36 = 5.16, HF ε = 0.99, p < 0.01)nd task by laterality (F2,36 = 6.24, HF ε = 1.0, p < 0.01), showeds, for this time window, orthographic task was significativelyore negative over anterior locations, whereas phonological and

emantic tasks activated a more distributed network, involvingoth anterior and posterior regions. Moreover, only orthographicnd phonological tasks were significantly left lateralized, while

emantic task exhibited a bilateral activation. These findingsere further clarified by the significant three-way task by regiony laterality interaction (F2,36 = 5.02, HF ε = 0.84, p < 0.01; seeig. 4a). Orthographic and phonological tasks showed a similar

C. Spironelli, A. Angrilli / Behavioural Brain Research 175 (2006) 296–304 299

espon

proor

pm

Fn

Fig. 2. Behavioral responses to W2. (a) R

attern: in fact both elicited a potential left-lateralized over ante-ior (p < 0.001) and posterior (p < 0.01) locations. However, the

rthographic task induced also a significant greater negativityver left anterior than posterior sites (p < 0.01) and a significanteduced positivity over right anterior compared with ipsilateral

rp(

ig. 3. Spline maps of orthographic (a and d), phonological (b and e) and semantic taegativity-activation; red, positivity-inhibition.

se times in milliseconds; (b) error rates.

osterior sites (p < 0.01). Instead, phonological task evoked aore extensive negativity over left hemisphere, as both ante-

ior and posterior locations were equally activated, and greaterositivity over right anterior compared with right posterior sitesp < 0.05). Semantic task, instead, evidenced a bilateral activa-

sks (c and f) during the iCNV (a–c) and the tCNV (d–f). Blue indicates cortical

300 C. Spironelli, A. Angrilli / Behavioural Brain Research 175 (2006) 296–304

F . (a) Tb durin

tt

earivLiau

gcrwTitr

Fa

ig. 4. Electrocortical responses of subjects during the three analyzed intervalsy region by laterality interaction during iCNV; (c) task by laterality interaction

ion pattern, with significant greater negativity over posteriorhan anterior locations of the right hemisphere (p < 0.01).

The second time interval (1–2 s) showed significant mainffects of both region factor (F1,18 = 5.33, p < 0.05), that revealed

significant greater negativity over posterior than anterioregions (−0.32 �V versus 0.05 �V, respectively), and lateral-ty factor (F1,18 = 11.56, p < 0.01), which evidenced greater leftersus right negativity (−0.45 �V versus 0.18 �V, respectively).

ike in the first time interval, we found the same significant

nteractions task by region (F2,36 = 6.10, HF ε = 0.80, p < 0.01)nd task by laterality (F2,36 = 7.79, HF ε = 1.00, p < 0.001). Butnlike prior time window, the semantic task was associated with

oit

ig. 5. Grand-averages obtained from all subjects: comparison between orthographiccross 38 channels. Negativity is upward.

ask by region by laterality interaction during the W1 (0.5–1 s) interval; (b) taskg tCNV.

reater negativity over posterior locations, whereas phonologi-al and orthographic tasks activated both anterior and posterioregions. Also in this epoch, orthographic and phonological tasksere left lateralized, and semantic task was activated bilaterally.he significant three-way interaction task by region by lateral-

ty (F2,36 = 3.37, HF ε = 0.88, p < 0.05; see Fig. 4b) summarizedhese findings and further clarified the development of the task-elated elaboration (Fig. 3a–c).

Thus, orthographic task was significantly left lateralized bothver anterior (p < 0.01) and posterior (p < 0.05) locations, show-ng left negativity and right positivity (see Fig. 3a). Phonologicalask appeared left lateralized too, at both anterior (p < 0.001) and

(light grey line), phonological (dark grey line) and semantic (black line) tasks

ral B

phptpSta

nwlt(cFa

3

wtApmwsvec−

4

bpugwactcCbl

plagiaoa

d[tTrj

aasrasarttbtsiitgaiacocaofcirapidcsni(sptfaoft

C. Spironelli, A. Angrilli / Behaviou

osterior sites (p < 0.01), exhibiting greater negativity on leftemisphere and greater positivity of right anterior versus rightosterior sites (p < 0.001; see Fig. 3b). Similarly to the previousime interval, we found greater right anterior positivity for thehonological with respect to the orthographic task (p < 0.001).emantic task showed, also in this interval, a bilateral activa-

ion, with relative negativity at posterior and marked positivityt anterior clusters (p < 0.01; see Fig. 3c).

In the last time interval (2–3 s), ANOVA revealed a sig-ificant main effect of laterality (F1,18 = 5.14, p < 0.05) factor,hich confirmed the overall significant greater negativity of

eft than right hemisphere (−0.60 �V versus −0.10 �V, respec-ively). The significant two-way task by laterality interactionF2,36 = 3.45, ε = 0.90, p < 0.05; see Fig. 4c) pointed to a signifi-ant left negativity for the phonological task only (p < 0.01; seeig. 3e), and a bilateral activation for the other two tasks (Fig. 3dnd f).

.3. CNV

Concerning the central cluster of electrodes (see Fig. 1b),e analyzed the three intervals to evaluate the temporal evolu-

ion of the central motor CNV. During the first window (0.5–1 s),NOVA revealed only the main effect of sex factor (F1,18 = 5.85,< 0.05), showing a greater cortical positivity in females than inales. During the second interval (1–2 s), no significant effectas found. Instead, task was the only factor that reached the

ignificance (F2,36 = 5.90, HF ε = 1.00, p < 0.01) in the last inter-al (2–3 s), in correspondence of which the orthographic tasklicited a significant greater negativity than both the phonologi-al (p < 0.05) and the semantic (p < 0.01) ones (−2.38 �V versus1.54 �V versus −1.25 �V, respectively; see Fig. 5).

. Discussion

Slow evoked potentials are quite large potentials generatedy wide superficial cortical layers [13,49] and, unlike earlyotentials (P200, P300, N400), their negativity is a relativelynambiguous index of cortical activation [13,49]. The contin-ent negative variation (CNV) is reduced in Alzheimer disease,here cortical atrophy is dominant, and is strongly inhibited by

nticholinergic drugs acting on cortical layers (see [13] for aritical review). Recent studies have further demonstrated thathe CNV is selectively inhibited/reversed in correspondence ofortical lesions [5,6]. We carried out this study to validate ourNV paradigm [4–6,40] as a non-invasive tool for analyzingoth the spatial aspects and the temporal dimension of linguisticateralization across different linguistic tasks.

The first important finding regarded the replication of ourrevious work, in which only the whole ISI interval was ana-yzed. In fact, in the past study [4] we found the same significantsymmetrical pattern for the phonological task which elicitedreater activation of left anterior regions. A number of past stud-

es has provided converging evidence, both with metabolic (PETnd fMRI) [17,19,27,42,67,69] and electrophysiological meth-ds [3,4,53–55,65], that phonological processes elicited withvariety of tasks always activate left anterior cortex. More in

idcs

rain Research 175 (2006) 296–304 301

etail, looking at our specific phonological task, Poldrack et al.45] used a rhyming judgement task with pseudowords to iden-ify the cortical regions involved in the phonological processing.he authors found that a significant activation of the left infe-

ior frontal cortex marked specifically the phonological rhymeudgement process.

For the semantic task, in each time window we found, alongntero-posterior regions, a bilateral activation that differentiatedcross time. In fact, during the first word analysis (0.5–1 s),ubjects exhibited a clear greater right posterior versus ante-ior activity, a result that points to the critical involvement ofcategorization network [1,4,44,58,66], and lexical–semantic

torages more distributed across hemispheres over posteriorreas. The increased bilateral activity over posterior brainegions during the two following time intervals, correspondingo word encoding in lexical–semantic networks (iCNV) and tohe response preparation after word rehearsal (tCNV), seems toe in agreement with previous results [4,66]. The fMRI litera-ure concerning the cortical structures involved in semantic taskseems to conflict in general with most EEG studies, and specif-cally with our findings, since many past studies with metabolicndices pointed to a significant involvement of left IFC in theseasks (see [17] for a review). In general, metabolic studies show areater activation of left hemisphere in most linguistic tasks andre less sensitive than EEG-based studies to right hemispherenvolvement. In most literature on ERPs and semantic processes,bilateral activation has been consistently found [1,4,18,61]. Aareful analysis of the literature allowed us to find a numberf possible relevant factors playing a role in this consistent dis-repancy. First of all, paradigms used so far to elicit semanticctivation in fMRI–PET investigations were quite different fromur, as only a small number of experiments used the same wordsor different tasks. Moreover, most of the tasks that have beenlassified as semantic in the literature on metabolic brain activ-ty were essentially semantic categorization or semantic memoryetrieval tasks (see for a large review [17]), which are tasks rel-tively more complex involving and embedding a number ofrocesses not easy to be disentangled [32]. Two methodolog-cal issues are also relevant for the observed differences. Toate, no reliable relationship has been found between metabolichanges in nervous tissue and the gross electrical activity mea-ured from neurons. It is even more surprising that there iso reliable correlation between the two most used metabolicndices, fMRI and PET, the first being an indirect measurefrom hemoglobin deoxygenation) and the second a direct mea-ure of neurons’ metabolism, an argument that has been raisedrecisely to explain differences between studies in the localiza-ion of semantic elaboration in the human brain [15]. For theMRI–EEG difference, this can be more easily comprehended:change of electrical activity can be measured from a group

f neurons before a bulk change of oxygen–glucose is requiredrom the same active tissue. Thus, EEG can be more sensitiveo small changes of neuron’s activity, compared with metabolic

ndices, and an electrical cortical activity can be in principleetected in the right hemisphere without a parallel measurablehange of metabolism. In line with this, most of ERP studies onemantic elaboration found activity also on the right hemisphere

3 ral B

[attttivtwi(inttmm

fatcoawovpmlaipsr

bastfpitoitbrrtpojlt

rpttlatapFttrwftpdpt

cavwfpcdvoiwlohc

ssafpduct

R

02 C. Spironelli, A. Angrilli / Behaviou

1,4,18,61]. This result is in agreement also with research onphasic patients showing some semantic–lexical competence ofhe right hemisphere when replacing the corresponding deficit ofhe left hemisphere [8]. A further methodological issue concernshe timing of linguistic processes: the time intervals analyzed inhe present study represent a fraction of the interval consideredn fMRI research (1 s was the longest ERP interval analyzedersus 6 s in a typical BOLD fMRI signal). It is well possiblehat fMRI measures the activity of the whole linguistic networkithin the left hemisphere that might be active also during those

ntervals that in ERP studies are not analyzed nor consideredbaseline, inter-trial interval, response interval). Thus, networksn the left hemisphere that are activated during intervals that areeglected in ERP studies would weight more on the final statis-ic on semantic activity. However, since most ERP studies pointo a bilateral distribution of semantic activity it seems that the

ain explanation for the discrepancy with research made withetabolic measures is essentially methodological.The current study replicated our prior results [4] obtained

rom a young sample of students, in a new sample of adultsged about 60. Therefore, the linguistic hemispheric specializa-ion seems to be a very robust feature not sensitive to progressiveognitive decay. Indeed, in a developmental ERP study carriedut on a large sample of right-handed subjects aged between 7nd 23, Grossi et al. [30] found that the CNV over anterior sitesas clearly evident starting at the age of 13–14. In that study, thebserved increase in the CNV over anterior cortical sites duringisual rhyming judgement task, was left lateralized and resultedositively correlated both with subjects’ age, and with the auto-atic phonological processing: in fact, children with higher

inguistic performance scores exhibited larger anterior CNVsymmetry. Such findings demonstrate that linguistic special-zation is a process related to training in reading and writing androbably begins not earlier than the 11–12th year. The left hemi-phere dominance for linguistic functions seems to be the mostobust neural correlate of the automation of speech abilities.

In the present study, to better contrast the activity induced byoth the semantic and the phonological judgement tasks, we usedn additional control task. Thus, during the orthographic task,ubjects’ were asked to perform a visual word match supposedo involve both perceptual processes, and very basic linguisticunctions. On one hand, data evidenced that orthographic andhonological tasks elicited a comparable left-lateralized activ-ty, this result led us to hypothesize that the two tasks, at least athe beginning, share the same automatic phonological encodingf the word, and perhaps its next retrieval in the verbal work-ng memory (vWM) storage [62]. On the other hand, however,hese two tasks differed as the phonological was characterizedy a significant greater inhibition-positivity over right anterioregions in the two early intervals (in ERP experiments is theelative left–right potential difference which is relevant ratherhan the absolute potential recorded in one site). Furthermore,osterior brain regions were significantly less activated in the

rthographic but not in the phonological task: thus, for rhymeudgement, a more distributed linguistic negativity in the wholeeft hemisphere was involved. Indeed, even though to performhe orthographic task correctly, only a perceptual visual match is

rain Research 175 (2006) 296–304

equired, it is probable that an automatic word elaboration tooklace, mainly in the first phases, as marked by the left lateraliza-ion. Finally, during the terminal CNV interval (2–3 s), whereashe phonological task showed a sustained and consistent left-ateralized activation, the orthographic task showed a bilateralctivation. We interpret this result as indicative of a shift fromhe automatic word processing to the specific requirement ofvisual word matching task, the most specific non-lateralized

rocess expected to be involved in the orthographic judgement.urther evidence in favor of differentiated processes underlying

he three tasks resulted from CNV analysis of central elec-rodes, particularly during the last time interval preceding motoresponse preparation [13,49,51]. In fact, in the tCNV conditione found a significant greater negativity over central electrodes

or the orthographic with respect to both the phonological andhe semantic tasks, a result pointing to an earlier motor responsereparation specifically for this task. This finding seems to vali-ate the idea that, at least during the last time interval, the wordrocessing was probably completed earlier for the orthographicask, whereas it was still in progress in the other two tasks.

In the present experiment, some degree of overlap among theortical networks involved in the selected tasks was expecteds the use of the same set of words in all tasks tends to acti-ate always the same nodes of automatic processing of theseords (lexical, semantic, phonological . . .), nevertheless, this

eature of the paradigm plays against our hypothesis of a sup-osed difference between the tasks, and any significant effectan be attributed to task manipulation only: for this reason anyifference between tasks represents a strong result. Statisticalalidity was further strengthened by the independent replicationf results from our prior experiment run on both German and Ital-an students. Another relevant aspect of the present experimentas represented by the possibility to follow the time course of the

inguistic processes engaged by the tasks. Without the analysisf three functionally different time windows, the tasks wouldave not been easily distinguished as they exhibited differentapability to elicit left lateralization mainly across intervals.

In conclusion, the present results support the view that hemi-pheric lateralization of different linguistic processes is con-istent across languages and across a wide range of subjectge. This feature makes the present paradigm particularly usefulor studying the lateralization of linguistic cortical networks inatients with language disorders [5,6,40], namely aphasics andyslexic children, thereby allowing to study the activity of resid-al networks survived to a lesion or the recruitment of alternativeompensatory networks involved in the plasticity associated tohe recovery of patient’s linguistic functions.

eferences

[1] Abdullaev YG, Posner MI. Time course of activating brain areas in gener-ating verbal associations. Psychol Sci 1997;8:56–9.

[2] Abdullaev YG, Posner MI. Event-related brain potential imaging of seman-

tic encoding during processing single words. NeuroImage 1998;7:1–13.

[3] Altenmuller E, Kriechbaum W, Helber U, Moini S, Dichgans J, PetersenD. Cortical dc-potentials in identification of the language-dominant hemi-sphere: linguistical and clinical aspects. Acta Neurochir 1993;56(SupplWien):20–33.

ral B

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

C. Spironelli, A. Angrilli / Behaviou

[4] Angrilli A, Dobel C, Rockstroh B, Stegagno L, Elbert T. EEG brain map-ping of phonological and semantic tasks in Italian and German languages.Clin Neurophysiol 2000;111:706–16.

[5] Angrilli A, Elbert T, Cusumano S, Stegagno S, Rockstroh B. Temporaldynamics of linguistic processes are reorganized in aphasics’ cortex: anEEG mapping study. NeuroImage 2003;20:657–66.

[6] Angrilli A, Spironelli C. Cortical plasticity of language measured by EEGin a case of anomic aphasia. Brain Lang 2005;95:32–3.

[7] Bauer H, Korunka C, Leodolter M. Technical requirements for high-quality scalp dc-recordings. Electroencephalogr Clin Neurophysiol1989;72:545–7.

[8] Benson DF. Aphasia and lateralization of language. Cortex 1986;22:71–86.[9] Benson RR, FitzGerald DB, LeSueur LL, Kennedy DN, Kwong KK,

Buchbinder BR, et al. Language dominance determined by whole brainfunctional MRI in patients with brain lesions. Neurology 1999;52:798–809.

10] Berg P, Scherg M. Dipoles models of eye movements and blinks. Elec-troencephalogr Clin Neurophysiol 1991;79:36–44.

11] Berg P, Scherg M. A multiple source approach to the correction of eyeartifacts. Electroencephalogr Clin Neurophysiol 1994;90:229–41.

12] Binder JR, Swanson SJ, Hammeke TA, Morris GL, Mueller WM, FischerM, et al. Determination of language dominance using functional MRI: acomparison with the Wada test. Neurology 1996;46:978–84.

13] Birbaumer N, Elbert T, Canavan A, Rockstroh B. Slow potentials of thecerebral cortex and behaviour. Physiol Rev 1990;70:1–41.

14] Bisiacchi P, Mondini S, Angrilli A, Marinelli K, Semenza C. Mass andcount nouns show distinct EEG cortical processes during an explicit seman-tic task. Brain Lang 2005;95:98–9.

15] Bookheimer S. Functional MRI of language: new approaches to under-standing the cortical organization of semantic processing. Annu Rev Neu-rosci 2002;25:151–88.

16] Bortolini V, Tagliavini C, Zampolli A. Lessico di frequenza della linguaitaliana contemporanea [Lexical frequency in current Italian]. Milano, Italy:Aldo Garzanti Editore; 1972.

17] Cabeza R, Nyberg L. Imaging cognition II: an empirical review of 275 PETand fMRI studies. J Cogn Neurosci 2000;12:1–47.

18] Cobianchi A, Giaquinto S. Event-related potentials to Italian spoken words.Electroencephalogr Clin Neurophysiol 1997;104:213–21.

19] Demonet J, Wise R, Franckowiak R. Language functions explored innormal subjects by positron emission tomography. Hum Brain Mapp1994;1:39–47.

20] Deppe M, Knecht M, Papke K, Lohmann H, Fleischer H, Heindel W, et al.Assessment of hemispheric language lateralization: a comparison betweenfMRI and fTCD. J Cereb Blood Flow Metab 2000;20:263–8.

21] Desmond J, Sum J, Wagner A, Demb J, Shear P, Glover G, et al. FunctionalMRI measurement of language lateralization in Wada-tested patients. Brain1995;118:1411–9.

22] Dobel C, Pulvermuller F, Harle M, Cohen R, Kobbel P, Schonle P-W, et al.Syntactic and semantic processing in the healty and aphasic human brain.Exp Brain Res 2001;140:77–85.

23] Dobel C, Cohen R, Berg P, Nagl W, Zobel E, Kobbel P, et al.Slow event-related brain activity of aphasic patients and controls inword comprehension and rhyming tasks. Psychophysiology 2002;39:747–58.

24] Elbert T, Dobel C, Angrilli A, Stegagno L, Rockstroh B. Word vs. taskrepresentation in neural networks. Behav Brain Sci 1999;22:286–7.

25] Friederici AD, von Cramon YD, Kotz SA. Language-related brain poten-tials in patients with cortical and subcortical left hemisphere lesions. Brain1999;122:1033–47.

26] Friederici AD, Alter K. Lateralization of auditory language functions: adynamic dual pathway model. Brain Lang 2004;89:267–76.

27] Frith CD, Friston KJ, Liddle PF, Frackowiak RS. A PET study of wordfinding. Neuropsychologia 1991;29:1137–48.

28] Geschwind N, Galaburda AM. Cerebral lateralization. Biological mech-anisms, associations, and pathology. II. A hypothesis and a program forresearch. Arch Neurol 1985;42:521–52.

29] Gratton G, Coles MGH, Donchin E. A new method for off-line removal ofocular artifact. Electroencephalogr Clin Neurophysiol 1983;55:468–84.

[

[

rain Research 175 (2006) 296–304 303

30] Grossi G, Coch D, Coffey-Corina S, Holcomb PJ, Neville HJ. Phonologicalprocessing in visual rhyming: a developmental ERP study. J Cogn Neurosci2001;13:610–25.

31] Hagoort P, Wassenaar M, Brown C. Real-time semantic compensationin patients with agrammatic comprehension: electrophysiological evi-dence for multiple-route plasticity. Proc Natl Acad Sci USA 2003;100:4340–5.

32] Hickok G, Poeppel D. Dorsal and ventral stream: a framework forunderstanding aspects of the functional anatomy of language. Cognition2004;92:67–99.

33] Hiscock M. Brain lateralization across the life span. In: Stemmer B,Whitaker HA, editors. Handbook of neurolinguistics. San Diego: AcademicPress; 1998. p. 357–68.

34] Holcomb PJ, Neville HJ. Natural speech processing: event-related brainpotentials. Psychobiology 1991;19:286–300.

35] Huynh H, Feldt LS. Conditions under which mean square ratios inrepeated measurements designs have exact F-distribution. J Am Stat Assoc1970;65:1582–9.

36] Liotti M, Gay CT, Fox PT. Functional imaging and language: evidence frompositron emission tomography. J Clin Neurophysiol 1994;11:175–90.

37] Luria AR. The working brain. New York: Basic Books; 1973.38] Oldfield RC. The assessment and analysis of handedness: the Edinburgh

inventory. Neuropsychologia 1971;9:97–113.39] Oostenveld R, Praamstra P. The five percent electrode system for high-

resolution EEG and ERP measurements. Clin Neurophysiol 2001;112:713–9.

40] Penolazzi B, Spironelli C, Vio C, Angrilli A. Altered hemispheric asymme-try during word processing in dyslexic children: an event-related potentialstudy. Neuroreport 2006;17:429–33.

41] Perrin F, Pernier J, Bertrand O, Echallier JF. Spherical splines for scalppotential and current density mapping. Electroencephalogr Clin Neuro-physiol 1989;72:184–7.

42] Petersen SE, Fox PT, Posner MI, Mintun M, Raichle ME. Positron emissiontomographic studies of the cortical anatomy of single-word processing.Nature 1988;331:585–9.

43] Poeppel D. A critical review of PET studies of phonological processing.Brain Lang 1996;55:317–51.

44] Poldrack RA, Wagner AD, Prull MW, Desmond JE, Glover GH, GabrieliJDE. Functional specialization for semantic and phonological processingin the left inferior prefrontal cortex. NeuroImage 1999;10:15–35.

45] Poldrack RA, Temple E, Protopapas A, Nagarajan S, Tallal P, MerzenichM, et al. Relations between the neural bases of dynamic auditory process-ing and phonological processing: evidence from fMRI. J Cogn Neurosci2001;13:687–97.

46] Posner MI, Pavese A. Anatomy of word and sentence meaning. Proc NatlAcad Sci USA 1998;95:899–905.

47] Pulvermuller F. Words in the brain’s language. Behav Brain Sci1999;22:253–336.

48] Pulvermuller F, Lutzenberger W, Preissl H. Nouns and verbs in the intactbrain: evidence from event-related potentials and high-frequency corticalresponses. Cereb Cortex 1999;9:497–506.

49] Rockstroh B, Elbert T, Canavan A, Lutzenberger W, Birbaumer N. Slowcortical potentials and behaviour. Munchen: Urban und Schwarzenberg;1989.

50] Roland PE, Lassen NA, Friberg L. Regional cortical blood flow changesduring fluent speech in subjects with verified hemispheric dominance. JCereb Blood Flow Matab 1985;5:205–6.

51] Rosler F, Heil M, Roder B. Slow negative brain potentials as reflections ofspecific modular resources of cognition. Biol Psychol 1997;45:109–42.

52] Ruchkin DS, Johnson R, Maheffey D, Sutton S. Toward a functional cate-gorization of slow waves. Psychophysiology 1988;25:339–53.

53] Ruchkin DS, Johnson Jr R, Grafman J, Canoune H, Ritter W. Multiple visu-ospatial working memory buffers: evidence from spatiotemporal patterns

of brain activity. Neuropsychologia 1997;35:195–209.

54] Rugg MD. Event-related potentials and the phonological processing ofwords and non-words. Neuropsychologia 1984;22:435–43.

55] Rugg MD. Event-related potentials in phonological matching tasks. BrainLang 1984;23:225–40.

3 ral B

[

[

[

[

[

[

[

[

[

[

[

[

04 C. Spironelli, A. Angrilli / Behaviou

56] Rutten GJ, Ramsey NF, van Rijen PC, Alpherts WC, van Veelen CW. fMRI-determined language lateralization in patients with unilateral or mixedlanguage dominance according to the Wada test. NeuroImage 2002;17:447–60.

57] Sartori G, Lombardi L. Semantic relevance and semantic disorders. J CognNeurosci 2004;16:439–52.

58] Seghier ML, Lazeyras F, Pegna AJ, Annoni JM, Zimine I, Mayer E, etal. Variability of the fMRI activation during a phonological and semanticlanguage task in healthy subjects. Hum Brain Mapp 2004;23:140–55.

59] Semlitsch HV, Anderer P, Schuster P, Presslich O. A solution for reliableand valid reduction of ocular artifacts applied to the P300 ERP. Psychophys-iology 1986;23:695–703.

60] Shaywitz BA, Shaywtiz SE, Pugh KR, Constable RT, Skudlarski P, Ful-bright RK, et al. Sex differences in the functional organization of the brainfor language. Nature 1995;373:607–9.

61] Skrandies W. Evoked potential correlates of semantic meaning—a brain

mapping study. Cogn Brain Res 1998;6:173–83.

62] Smith EE, Jonides J. Storage and executive process in the frontal lobes.Science 1999;283:1657–61.

63] ter Keurs M, Brown CM, Hagoort P, Stegeman DF. Electrophysiologicalmanifestation of open- and closed-class words in patients with Broca’s

[

[

rain Research 175 (2006) 296–304

aphasia with agrammatic comprehension. An event-related brain potentialstudy. Brain 1999;122:839–54.

64] Thiel A, Herholz K, von Stockhausen HM, Leyen-Pilgram K, PietrzykU, Kessler J, et al. Localization of language-related cortex with 15O-labeled water PET in patients with gliomas. NeuroImage 1998;7:284–95.

65] Thomas C, Altenmuller E, Marckmann G, Kahrs J, Dichgans J. Languageprocessing in aphasia: changes in lateralization patterns during recoveryreflect cerebral plasticity in adults. Electroencephalogr Clin Neurophysiol1997;102:86–97.

66] Thompson-Schill SL, D’Esposito M, Aguirre GK, Farah MJ. Role of leftinferior prefrontal cortex in retrieval of semantic knowledge: a reevaluation.Proc Natl Acad Sci USA 1997;94:14792–7.

67] Warburton E, Wise RJ, Price CJ, Weiller C, Hadar U, Ramsay S, et al.Noun and verb retrieval by normal subjects. Studies with PET. Brain1996;119:159–79.

68] Yetkin FZ, Swanson S, Fischer M, Akansel G, Morris G, Mueller W, et al.Functional MR of frontal lobe activation: comparison with Wada languageresults. Am J Neuroradiol 1998;19:1095–8.

69] Zatorre RJ, Evans AC, Meyer E, Gjedde A. Lateralization of phonetic andpitch discrimination in speech processing. Science 1992;256:846–9.