An Electrophysiologic Study: Can Seman tic Con text Pro cesses Be Mobi lized in Patients With Thought-Disordered Schizo phre nia?

Original Research An Electrophysiologic Study: Can Seman tic Con text Pro cesses Be Mobi lized in Patients With Thought-Disordered Schizo phre nia? Milena Kostova, PhD 1, Chris tine Passerieux, MD, PhD 2, Jean-Paul Laurent, PhD 3, Ma rie-christine Hardy-Baylé, MD, PhD 4 Objective: The aim of this study was to test the hy poth e sis that re in forc ing the struc tur ing of ver bal ma te rial may bring about an im prove ment in contextualization pro cesses in pa - tients with for mal thought-disordered schizo phre nia. Method: A to tal of 38 schizo phre nia pa tients with for mal thought dis or ders and 24 matched healthy con trol sub jects per formed 2 lex i cal de ci sion tasks, in volv ing 2 lev els of con tex tual struc tur ing (with 16.7% and 33% of re lated words, re spec tively). The eventre lated po ten tials, N400 and late pos i tive com po nent (LPC), and be hav ioural vari ables ( reaction times and er ror per cent ages) were an a lyzed. Re sults: A con text-structuring ef fect was ob served on LPC, but not on N400. In sub jects with schizo phre nia, the N400 anom a lies (that is, in crease in am pli tude for the re lated words and re duc tion of the N400 ef fect) per sisted in both con text-structuring con di tions. Sim i larly, a re duc tion in LPC am pli tude for the un re lated word cat e gory, as well as a de - crease in the LPC ef fect, was ob served in these pa tients. Con clu sions: The schizo phre nia pa tients with for mal thought dis or ders did not ben e fit from the struc tur ing of the con text to im ple ment con text in te gra tion strat e gies. This def i cit ap pears to be sta ble. The re sults are dis cussed within the frame work of a pre vi ously published model of lan guage com pre hen sion. (Can J Psy chi a try 2003;48:615 623) In for ma tion on au thor af fil i a tions ap pears at the end of the ar ti cle. Clin i cal Im pli ca tions The im pair ment of the in te gra tive pro cesses that are in volved in the pro cess ing of con text is con firmed in pa tients with thought-disordered schizo phre nia. This def i cit ap pears to be sta ble and does not ben e fit from any cog ni tive remediation. The hy poth e sis pro posed in this ar ti cle that the pre dic tive pro cesses are pre served while the in te gra tive contextualization pro cesses are im paired makes ef forts at cog ni tive re ha bil i ta tion con ceiv able. Lim i ta tions The pro posed hy poth e sis; namely, that there is a def i cit in the in te gra tive con text-pro cess ing strat e gies in schizo phre nia patients and that the pre dic tive strat e gies are pre served will have to be tested di rectly. The ecologic na ture of the in ves ti gated pro cesses will have to be con firmed in sit u a tions closer to that of nat u ral lan guage (for ex am ple, us ing a phrasal con text). Key Words: schizo phre nia, thought dis or ders, lan guage, con text pro cess ing, N400, late pos i tive com po nent Can J Psy chi a try, Vol 48, No 9, Oc to ber 2003 615

The Ca na dian Jour nal of Psy chia try Orig i nal Re search The def i cit in se man tic con text use has been reg u larly iden - ti fied in nu mer ous stud ies of lan guage in schizo phre nia patients, and its psychopathological rel e vance to the study of thought and com mu ni ca tion dis or ders in schizo phre nia has been stressed on many oc ca sions. Thanks to their high level of tem po ral res o lu tion, the event-related poten tials (ERPs) simul ta neously pro vide research ers with an accu rate mea sure of on-line lan guage pro - cess ing and a direct man i fes ta tion of brain activ ity. The N400 com po nent appears approx i mately 250 milliseconds to 500 milliseconds after the pre sen ta tion of a poten tially sig nif i cant stimulus (1). Further, its amplitude varies systematically dur - ing seman tic pro cess ing and can be reduced by var i ous fac tors that increase an item s pre dict abil ity within the local con text or by the pres ence of seman tic links between the words. The N400 ampli tude is extremely sen si tive to the imme di ate con text in which the word appears, whether this con text con - sists of a word or a sen tence (2). Gen erally, N400 reflects the inte gra tion of the con text: inte gra tion is eas ier and N400 ampli tude is lower when the word is con gru ent with the con - text ( N400 effect ). Like wise, it has been dem on strated that N400 is sen si tive to the ease with which sub jects can access the infor ma tion stored in seman tic mem ory, inde pend ently of the plau si bil ity and the pre dict abil ity of the item within the local con text in rela tion to the seman tic links between the words (3). N400 is habit u ally fol lowed by a late pos i tive com po nent (LPC) which, like N400, has proved to be sen si tive to seman - tic con gru ence (4,5). This effect, how ever, exhib its an inverse polar ity: incon gru ent words evoke a more highly pos i tive poten tial than do con gru ent words. Although the func tional sig nif i cance of LPC and its rela tion with N400 have not yet been clearly iden ti fied, the pre dom i nant opin ion is that LPC reflects an addi tional stage in the pro cess ing of the stim u lus, occur ring after eval u a tion and before the choice of the response, that is related to the updat ing of work ing mem ory (6). In a some what sim i lar vein, Van Petten and oth ers con - sider that LPC reflects a more elabo ra tive pro cess ing of the stim u lus and its inte gra tion in mem ory (5). Finally, for Halgren, N400 and LPC reflect dif fer ent stages in the same cog ni tive pro cess (7). The influ ence of the seman tic con text on word pro cess ing can be stud ied using the lex i cal deci sion par a digm. In this task, sub jects decide whether or not the tar get word exists in their lan guage. The pres ence of a seman tic or asso cia tive link between the tar get and the prime word that pre cedes it has been sys tem at i cally found to reduce the time required to make this lex i cal deci sion. This phe nom e non has been termed seman tic prim ing (8). The 3 electrophysiologic stud ies that have used the lex i cal deci sion in sub jects with schizo phre nia 616 report a nor mal N400 effect in these patients (9 11). These results are par tic u larly unex pected because they con tra dict most behav ioural stud ies (12), and because almost all N400 stud ies that have used seman tic judge ment tasks in a sen tence con text have found an anom aly of this ERP com po nent (reduc tion or absence of the N400 effect and [or] an increase in ampli tude in the con gru ent or related con di tion) (4,13 22). What the 3 lex i cal deci sion stud ies have in com mon is the high level of struc tur ing of the seman tic con text, with the over all mate rial con tain ing over 30% related words. The absence of a def i cit in schizo phre nia patients could there fore be explained through the mobi li za tion of con text-pro cess ing strat e gies, owing to the highly struc tured ver bal mate rial used in these stud ies. At the same time, behav ioural stud ies have clearly estab lished that the seman tic prim ing anom aly spe cif i cally con cerns patients with thought-disordered schizo phre nia (23). Sev eral electrophysiologic stud ies have also revealed a cor re la tion between for mal thought dis or ders and N400 anom a lies (20), or even between such dis or ders and N400 and LPC anom a lies con jointly (24). This clin i cal vari able, how ever, was not con - sid ered in the stud ies that used lex i cal deci sion. Finally, the stud ies that have inves ti gated LPC in patients with schizo - phre nia have iden ti fied anom a lies in this com po nent, which have taken the form of a gen eral reduc tion in ampli tude and the con gru ence effect (4,16,20,22,24). To pro vide a more finely honed func tional anal y sis of the pro cesses that were mobi lized in our exper i ments, we con sider the data relat ing to LPC. Thus, the main aim of the cur rent study was to com pare the behav ioural and electrophysiologic results obtained from a group of schizo phre nia patients with for mal thought dis or ders in a lex i cal deci sion task, using dif fer ent lev els of struc tur ing of the con text. In light of the data reported in the lit er a ture, we hypoth e sized that the rein force ment of the con tex tual con - straints (high pro por tion of related words) should allow patients with schizo phre nia to mobi lize seman tic con textpro cess ing strat e gies, which should be evi denced by the seman tic prim ing effect and an N400 effect, com pa ra ble with those observed in a group of healthy con trol sub jects. Conversely, in a weakly struc tured lex i cal deci sion task (low pro por tion of related words), the group of schizo phre nia patients should be char ac ter ized by the absence of seman tic prim ing and by N400 anom a lies. Methods and Mate rials Sub jects A total of 38 patients (includ ing 10 women) were recruited from Pr JF Che va lier s Psy chi at ric Depart ment at Ver sailles Hos pi tal. All sub jects were char ac ter ized by a period of Can J Psy chia try, Vol 48, No 9, October 2003

An Electrophysiologic Study: Can Seman tic Con text Pro cesses Be Mobi lized in Patients With Thought-Disordered Schizo phre nia? Table 1 Demographic, cognitive, and clinical data for the subgroups as a function of the experimental conditions Schizo phre nia pa tients Highly struc tured n = 19 Mean (SD) Weakly struc tured n = 19 Mean (SD) symptomatic stability. Psychotic symptomatology had been eval u ated using the Pos i tive and Neg a tive Syn drome Scale (PANSS) (25). The for mal thought dis or ders were eval u ated using the Thought, Lan guage, and Com mu ni ca tion Dis or ders Scale (TLCDS) (26). All patients were receiv ing neuroleptic treat ment, and the daily dosage was con verted into a chlorpromazine equiv a lent (27). The con trol group con sisted of 24 healthy con trol sub jects (includ ing 15 women), who were matched on age, aca demic level, and vocab u lary level with the patients who suf fered from schizo phre nia (28). All sub jects spoke French as their native lan guage and had nor mal or cor rected-to-normal vision. The exclu sion cri te ria for the 2 groups of sub jects were as fol lows: age under 20 years or over 50 years, neu ro log i cal ante ced ents, alco holic intox i ca tion, or reg u lar or recent use of toxics. Highly struc tured n = 12 Mean (SD) Con trol sub jects Weakly struc tured n = 12 Mean (SD) Age 31.28 (8.98) 30.11 (8.27) 29.92 (6.63) 30.92 (8.72) Education (years) 12.61 (3.48) 12.68 (3.09) 12.00 (2.63) 12.67 (2.53) Vocabulary 27.22 (6.56) 24.37 (6.40) 26.33 (5.52) 26.42 (4.76) PANSS total score 85.26 (16.73) 79.79 (21.45) PANSS Positive scale PANSS Negative scale 21.63 ( 6.22) 20.26 (7.15) 23.16 (4.84) 21.32 (6.26) TLCDS 14.56 (6.18) 13.32 (7.38) Neuroleptics a 725.05 (506.47) 673.23 (442.11) PANSS = Pos i tive and Neg a tive Syn drome Scale; TLCDS = Thought, Lan guage, and Com mu ni ca tion Dis or ders Scale a Chlorpromazine equivalent Table 2 Mean amplitude (microvolts) of the medial electrodes (Fz, Cz, and Pz) for the related and unrelated words in the 2 groups as a function of the level of structuring of the context 250 milliseconds to 450 milliseconds Schizo phre nia pa tients Con trol sub jects Highly struc tured Weakly struc tured Highly struc tured Weakly struc tured Re lated words 0.6 0.4 2.5 1.5 Un re lated words 0.0 0.8 0.4 0.4 450 milliseconds to 650 milliseconds Re lated words 1.0 1.6 0.5 2.0 Un re lated words 1.3 0.7 2.3 1.6 Stimuli The mate rial con sisted of 4 lists, each com - prising 300 pairs of items. Each list con - tained 150 word nonword pairs and 150 word word pairs. Of the lists, 2 were highly struc tured and con - tained 100 pairs of related words, includ - ing 50 pairs of fill ers (that is, the same characteristics but excluded from anal y - ses) and 50 pairs of unre lated words (that is, 33% of related words). The other 2 lists were weakly struc tured and con - tained 50 pairs of related words and 100 pairs of unre lated words, includ ing 50 pairs of fill ers (that is, 16.7% of related words). For each of these lists, the order of item presentation was fixed, fol low ing ran - dom iza tion. All sub - jects saw the same tar get words and the same nonwords; the lists were con structed by coun ter bal anc - ing the pre sen ta tion of the mate rial (that is, the tar get words that appeared in the related con di tion in one list appeared in the unre lated con di tion in the other, and vice versa). The type of seman tic link within the pairs con sisted of an ant - onym rela tion or mem ber ship of the same seman tic cat e gory, or con sti tuted a part or whole rela tion. All words were common, and the word length var ied between 3 and 7 let ters (mean 5 let ters per word). To elim i nate any bias because of the way the lists were con structed, the mate rial was tested on 10 healthy sub jects who did not take part in the study. Two train - ing lists, with 90 items each, were con structed on the basis of the same rules. Pro ce dure Of the 2 struc tur ing con di tions, 1 was ran domly assigned to the sub jects. A total of 19 patients and 12 con trol sub jects Can J Psy chi a try, Vol 48, No 9, Oc to ber 2003 617

The Ca na dian Jour nal of Psy chia try Orig i nal Re search Table 3 Rao R values of the MANOVAs for N400 and LPC at the medial sites (Fz, Cz, and Pz) were tested in each of the exper i men tal con di tions. The patients and the con trol sub jects in each con di tion were com - pared on the basis of the demo graphic, cog ni tive, and clin i cal criteria (Table 1). All sub jects were informed of the gen eral aims of the study and had signed a writ ten con sent form. They were seated com fort ably, approx i mately 80 centimetres in front of a com puter screen. They were told that they would see 2 sequences of let ters and that they had to decide as quickly and accu rately as pos si ble whether the sec ond sequence of let - ters was or was not a French word. They gave their response using their dom i nant hand, posi tioned on a 2-but ton mouse. The first stim u lus was pre sented in black, lower-case char ac - ters against the white com puter screen for a period of 200 milli seconds; the screen then became white again for 250 millisec onds before the sec ond stim u lus was dis played for 1200 mil li sec onds (stim u lus-onset asynchrony [SOA] of 450 milliseconds, interstimulus inter val [ISI] of 2000 mil li - sec onds ). The sub jects took part in a train ing ses sion before per form ing the task. The total record ing period was approx i - mately 30 min utes. ERP Re cord ing and Anal y sis To record the EEG, 12 elec trodes were arranged on the scalp accord ing to stan dard inter na tional prac tice: 3 elec trodes in the fron tal region (F3, Fz, and F4), 3 in the cen tral region (C3, Cz, and C4), 3 in the pari etal region (P3, Pz, and P4), 2 in the left (T3) and right (T4) tem po ral regions, and 1 in the occip i tal zone (Oz). We used 4 elec trodes to record the elec trooculogram (EOG): 2 at the level of the exter nal canthi and 1 above and 1 below the eye. All the imped ances were held below 1.8 kiloohms. The EEG was con tin u ously recorded using the InstEP sys tem, with a fre quency of 512 points per sec ond, and the eye move ments were then cor rected off-line using an auto matic pro gram. Finally, the data were dig i tally fil tered using a bandpass of 0.80 hertz to 12 hertz. Aver age wave forms were cal cu lated sep a rately for each sub - ject and stim u lus type with ref er ence to the 200-mil li sec ond prestimulus base line. The data were ana lyzed by cal cu lat ing, sep a rately for each sub ject and record ing site, the mean ERP ampli tude for each stim u lus type over two 200-mil li sec ond epochs: 1) 250 mil li sec onds to 450 mil li sec onds after stim u lus 618 Effect df Rao R P Relatedness 2,57 21.54 < 0.0001 Electrode 4,55 14.08 < 0.0001 Group relatedness 2,57 7.80 < 0.001 Structuring relatedness 2,57 11.17 < 0.0001 df = de grees of free dom; LPC = late pos i tive com po nent; MANOVA = muultivariate anal y sis of variance onset for N400 and 2) 450 mil li sec onds to 650 mil li sec onds after stim u lus onset for LPC. These epochs were selected after visual inspec tion of the wave forms. The response times and error per cent ages were ana lyzed sep a - rately using anal y ses of vari ance (ANOVAs), with group and struc tur ing as inter group fac tors and relat ed ness as intragroup fac tor. The ERP data anal y ses focused on the midline record - ing site (Fz, Cz, and Pz). Multivariate anal y ses of vari ance (MANOVAs) were per formed using group and struc tur ing as inter group fac tors and relat ed ness and elec trode site as intragroup factors. The significant interactions were analyzed sep a rately using post hoc Scheffé tests. Results Be hav ioural re sults The results of the response time anal y ses reveal a main effect of group (F 1,58 = 25.61; P < 0.0001), which indi cates that response times were gen er ally lon ger in schizo phre nia patients (774 mil li sec onds) than in con trol sub jects (619 mil li - sec onds), together with an effect of relat ed ness (F 1,58 = 23.53; P < 0.0001), with response times shorter for related (684 milli - seconds) than for unre lated words (709 mil li sec onds). Likewise, the anal y sis of the error per cent ages reveals an effect of group (F 1,58 = 17.34; P < 0.0001), with the error per - cent age sig nif i cantly higher among the sub jects who had schizo phre nia (13.42%) than among the con trol sub jects (2.67%). No effect of struc tur ing or any inter ac tion with this fac tor was observed. Electrophysiologic Re sults Fig ures 1a, 1b, 2a, and 2b pres ent the aver age wave forms for each task and stim u lus type for the 2 diag nos tic groups sep a - rately. Table 2 indi cates the mean ampli tudes for the midline record ing sites for the N400 and LPC epochs. Table 3 summarizes the main results of data anal y ses. The results indi cate the pres ence of a main effect of seman tic rela tion. Anal y ses of this effect reveal that, in the N400 epoch, the ampli tude of the related words is more pos i tive than that of the unre lated words (related word, 1.3 microvolts; unre lated words, 0.2 microvolts; F 1,58 = 37.99; P < 0.0001). We also observed a main effect of elec trode, explained by a dif fer ence in ampli tude between the elec trodes in the LPC win dow (F 2,116 = 29.68; P < 0.0001) with max i mum pos i tive ampli tudes at Pz (2.1 microvolts), com pared with Cz (1.6 micro volts; P < 0.04), and at Cz, com pared with Fz (0.4 microvolts; P < 0.0001). The group relatedness interaction was also significant. The anal y ses show that patients with schizo phre nia exhibit an increase in N400 ampli tude on related words, com pared with con trol sub jects (schizo phre nia patients, 0.5 microvolts; con - trol sub jects, 2.0 microvolts; F 1,58 = 6.15; P < 0.02) and a Can J Psy chia try, Vol 48, No 9, October 2003

An Electrophysiologic Study: Can Seman tic Con text Pro cesses Be Mobi lized in Patients With Thought-Disordered Schizo phre nia? Fig ure 1a ERPs for con trol sub jects in the highly struc tured task re lated words un re lated words ERPs = Event-re lated po ten tials; V = microvolts; ms = mil li sec onds Fig ure 1b ERPs for schizo phre nia pa tients in the highly struc tured task Difference Waveform Dif fer ence wave form, obtained after sub - tract ing the mean wave forms for the related words from those of the unre lated words (Fig ure 3), was cal cu lated for 9 elec - trodes: F3, Fz, and F4 for the fron tal region; C3, Cz, and C4 for the cen tral region; and P3, Pz, and P4 for the pari etal region. The MANOVAs were per formed using group and level of struc tur ing as inter group fac tors and topog ra phy (fron tal, cen tral, and pari etal) and elec trode (left, medial, and right) as intragroup fac tors. The results reveal a main effect of group (Rao R 2,57 = 8.32; P < 0.0007), which indi - cates a reduc tion in the ampli tude of the N400 effect (F 1,58 = 6.55; P < 0.01) and the LPC effect (F 1,58 = 7.05; P < 0.01) in patients who suf fer from schizo phre nia, com pared with con trol sub jects (schizo - phre nia patients, 0.7 microvolts and 0.2 microvolts, respec tively; con trol sub - jects: 1.8 microvolts and 0.6 microvolts, respectively). We also observed an effect of elec trode (Rao R 4,55 = 10.11; P < 0.0001), which indi - cates that the N400 effect is at its max i mum for the midline elec trodes ( 1.5 microvolts), com pared with the left hemi sphere ( 1.1 microvolts) and the right hemi sphere ( 1.1 microvolts) elec trodes. The ante rior and pos te rior dis tri bu tion of the effects did not dif fer. Dis cus sion re lated words un re lated words decrease in LPC ampli tude for unre lated words ( schizo - phrenia patients, 1.0 microvolts; con trol sub jects, 1.9 microvolts; F 1,58 = 6.49; P < 0.01). Finally, the anal y ses reveal the pres ence of a sig nif i cant struc - tur ing relat ed ness inter ac tion. The N400 effect is sig nif i cant in the 2 con text-structuring con di tions. In con trast, the ampli - tudes for related and unre lated words dif fer sig nif i cantly in the LPC win dow in the highly struc tured task only (related words, 0.8 microvolts; unre lated words, 1.8 microvolts; F 1,58 = 22.42; P < 0.0001). Owing to the absence of a main effect of struc tur ing or of any inter ac tion between this fac tor and the diag nos tic group, we will not con duct sep a rate anal y ses for each con textstructuring con di tion. Re sults for the Con trol Sub jects In accor dance with the data reported in the lit er a ture, we observed a mod u la tion in N400 ampli tude by the pres ence of a seman tic rela tion in a lex i cal deci sion task that involved dif fer ent lev els of con text struc tur ing. The manip u la tion of the pro por tion of related words did not affect the N400 effect, which exhib ited the same order of mag ni tude in the 2 lex i cal deci sion tasks. This result is sim i lar to that obtained by Silva-Pereyra and oth ers (29), who also failed to observe any effect of con text struc tur - ing on the N400 effect (13% vs 36% of related words). Con versely, we observed an effect of this struc tur ing on LPC: in the highly struc tured task, the unre lated words evoked more highly pos i tive ampli tudes than did related words, whereas in the weakly struc tured task, we observed a per sis tence of the N400 effect in the LPC win dow, with the related words evok - ing greater pos i tive ampli tudes. This result indi cates that the Can J Psy chi a try, Vol 48, No 9, Oc to ber 2003 619

The Ca na dian Jour nal of Psy chia try Orig i nal Re search Fig ure 2a ERPs for con trol subjects in the weakly struc tured task re lated words un re lated words Fig ure 2b ERPs for schizo phre nia patients in the weakly struc tured task re lated words un re lated words Fig ure 3 Dif fer ence wave form for con trol sub jects and schizo phre nia patients 2 lex i cal deci sion tasks did not mobi lize the same cog ni tive pro cesses. The highly struc tured task seems to have been pro - cessed at a deeper, more elabo ra tive level than the weakly struc tured task. Fur ther, the words that were unre lated to the con text were the object of addi tional inte gra tive pro cess ing. Re sults for Pa tients with Schizo phre nia The main result of our study is that N400 anom a lies per sist in schizo phre nia patients, even when the struc ture of the con text is enhanced. In fact, we observed a decrease in the N400 effect in this group of patients, owing to the evo ca tion of a high-amplitude N400 for the related words. The increase in N400 ampli tude for words that seman ti cally relate to the prime or that are con gru ent with the sen tence con text has been widely reported in the lit - er a ture (4,16 18,22), reflect ing a def i cit in the use of the con text pro vided by the pre - ced ing word (30). In this respect, our results con firm those pre vi ously reported in the literature. Con trary to our hypoth e ses, we observed N400 anom a lies not only in the weakly struc tured task but also in the highly struc - tured one. Thus, it appears that, in a group of patients with for mal thought-disordered schizo phre nia, the rein force ment of the con tex tual con straint, as operationalized in our study, does not per mit the mobi li za tion of con tex tual pro cesses. Like wise, it should be noted that an exam i na tion of ERPs reveals that the N400 anom a lies are lower in the weakly struc tured task, prob a - bly because of the more super fi cial level at which this task is pro cessed. 620 schizo phre nia patients con trol subjects The results obtained in our highly struc - tured task dif fer from those reported by Koyama and oth ers (10,11), even though the pro por tions of related words are com - pa ra ble. There are cer tain nota ble dif fer - ences, how ever, between these stud ies and our own first, con cern ing the clin i cal char ac ter is tics of the patients and sec ond, con cern ing the nature of the seman tic rela - tion and the SOA dura tion. Can J Psy chia try, Vol 48, No 9, October 2003

An Electrophysiologic Study: Can Seman tic Con text Pro cesses Be Mobi lized in Patients With Thought-Disordered Schizo phre nia? As we men tioned in our intro duc tion, the seman tic prim ing and N400 anom a lies do not seem to affect all schizo phre nia patients but only the sub group char ac ter ized by the pres ence of for mal thought dis or ders. The diver gence between our results and those obtained by Koyama and oth ers could be owing to the clin i cal het er o ge ne ity of the patients, itself under pinned by a cog ni tive het er o ge ne ity. Should we con clude that the def i cit in contextualization pro - cesses in schizo phre nia patients with for mal thought dis or ders is sta ble? Should we seek to explain those stud ies based on behav ioural vari ables that have reported com pa ra ble seman tic prim ing effects in schizo phre nia patients with for mal thought dis or ders and in con trol sub jects (31,32)? The func tional anal y sis of the exper i men tal tasks makes it pos si ble to advance a com ple men tary hypoth e sis: per haps, our highly struc tured task did not solicit the same type of con - text pro cesses as those mobi lized by stud ies in which a con text effect was reported. In fact, 3 types of mech a nism have been pro posed to explain the seman tic prim ing effect (33). These are the auto matic spread ing of acti va tion through the seman tic net work, a prelexical mech a nism for the gen er a tion of expec - ta tions, and a postlexical mech a nism for the ver i fi ca tion of seman tic com pat i bil ity. Accord ing to the clas sic oppo si tion of auto matic and con trolled pro cesses, the first of these mech a - nisms should be auto matic and be inde pend ent of the exper i - men tal con di tions, whereas the other 2 should be con trolled and closely linked to the char ac ter is tics of the mate rial and the task. In their ver bal mate rial, Koyama and oth ers used only anto - nymic rela tions. The preg nancy of the anto nymic rela - tions cou pled with the pre cise antic i pa tions that they per mit, the high pro por tion of related words, and the long SOA (greater than 1000 mil li sec onds) likely favoured the mobi li - za tion of the prelexical mech a nism for the gen er a tion of expec ta tions (34). This obser va tion also applies to the stud ies con ducted by Henik and oth ers (31,32) who employed all the con di tions that are con du cive to the expec ta tion gen er a tion mechanism. Con versely, our ver bal mate rial used var i ous seman tic rela - tions. Our exper i men tal con di tions, which employed a less preg nant seman tic rela tion and a shorter SOA (450 milliseconds), did not call on the expec ta tion-generation mech a - nism but instead on the postlexical-integration mech a nism, known to oper ate inde pend ently of SOA and to have effects that increase as a func tion of the pro por tion of related words (34). The per sis tence of N400 anom a lies under these con di - tions may indi cate a sta ble def i cit in the postlexical inte gra tive strat e gies in schizo phre nia patients with for mal thought dis or ders. The value of dis tin guish ing between prelexical pre dic tive strat e gies and postlexical inte gra tive strat e gies in a study of contextualization def i cits in schizo phre nia patients is also reflected in the gen eral lan guage com pre hen sion model recently pro posed by Kutas and Federmeier (35). Accord ing to this model, the pro cess ing of con text is thought to preactivate the perceptual and semantic characteristics of the most coher ent item or of the most highly pre dict able item within the given con text. Acti va tion then spreads through the seman tic net work to related con cepts, and when the pre sented item cor re sponds to the pre dic tions, it is rec og nized more eas - ily. When this proactive strat egy proves defec tive, the inte gra - tive strat egy is used. The sub ject needs to allo cate addi tional attentional resources to com pare the char ac ter is tics of the tar - get with those of the con text (that is, to search for an asso ci a - tion with the con text words or seman tic com pat i bil ity) and to inte grate the word into the rep re sen ta tion of the con text stored in work ing mem ory. It seems per mis si ble, then, to think that it is pre cisely this ret ro ac tive recourse to the con tents of work - ing mem ory and the addi tional cog ni tive effort it involves that is reflected in the LPC ampli tude increase fol low ing N400. A lit er a ture review on schizo phre nia patients, in light of this dual-process model, sug gests that, when ever it is pos si ble to use the pre dic tive strat egy, patients with schizo phre nia exhibit a prim ing effect that is equiv a lent to, or some times greater than, that observed in con trol sub jects. This is the case, for exam ple, in lex i cal deci sion tasks that use a highly struc - tured con text or in lex i cal deci sion tasks con tain ing ambig u - ous words whose nondominant mean ing is strongly induced by the con text. Sim i larly, when the exper i men tal con di tions do not neces si tate a recourse to the postlexical inte gra tive strat e gies (that is, pas sive read ing tasks), schizo phre nia patients do not exhibit N400 anom a lies (4,24). Nota bly, in our study, the LPC anom aly in the schizo phre nia patients was spe cific to the unre lated con di tion. (The same result was also obtained by Strandbourg and oth ers [22].) Taken together, the N400 anom a lies in the related con di tion only and the LPC in the unre lated con di tion only could indi - cate that sub jects with schizo phre nia, unlike the con trol sub - jects, are unable to use a seman tic rela tion to facil i tate word pro cess ing (that is, an increase in N400 ampli tude). Fur ther, these sub jects do not make the addi tional effort that is required to inte grate unre lated words (that is, a reduc tion in LPC amplitude). We can also observe that the ERP anom a lies seen in our study in indi vid u als with schizo phre nia do not appear to relate to neuroleptic treat ment. We wit nessed no cor re la tion between the daily dos age (mea sured in terms of chlorpromazine equiv - a lent) on the one hand and the N400 or LPC ampli tude on the other. This result is in line with those reported in sev eral dif - fer ent stud ies (14,18 20,24). More over, the only study pub - lished to date that com pares the N400 pat tern in a group of schizo phre nia patients receiv ing treat ment (haloperidol) with that of a group of untreated schizo phre nia patients and a healthy con trol group on a lex i cal deci sion task has revealed the pres ence of a sig nif i cant N400 effect in the con trol group and in the treated patient group, but not in the untreated patient group (9). These results indi cate that con ven tional neuroleptics actu ally tend to improve the effect of seman tic con text on N400 ampli tude in schizo phre nia patients. Conclusion The absence of any improve ment in the pro cess ing of con text in a con di tion that facil i tated the mobi li za tion of postlexical Can J Psy chi a try, Vol 48, No 9, Oc to ber 2003 621

The Ca na dian Jour nal of Psy chia try Orig i nal Re search inte gra tive strat e gies sug gests that the def i cit in these pro - cesses is sta ble in schizo phre nia patients. This def i cit may relate to work ing-memory anom a lies in schizo phre nia (20,21,31). These results, taken together with the absence of N400 anom a lies observed under con di tions that strongly solicit the pre dic tive strat e gies, lead us to believe that schizo - phre nia patients pro cess lan guage on the basis of the proactive antic i pa tions induced by the con text. They expe ri ence dif fi - cul ties, how ever, when they have to revise these expec ta tions via the ret ro ac tive recourse to the inte gra tion of mean ing. The weak en ing of the asso ci a tions, the loss of goal, and the asso ci - a tions based on asso nance or seman tic con ti gu ity (that is, the dis or ga ni za tion of lan guage and thought in schizo phre nia) could be the clin i cal expres sion of this cog ni tive dis or der. The hypothesis that the predictive strategies are preserved while only the inte gra tive strat e gies are impaired remains to be tested in later research. Ref er ences 1. Kutas M, Hillyard SA. Event-related brain po ten tials to se man ti cally in ap pro pri - ate and sur pris ingly large words. Biol Psychol 1980;11:99 116. 2. Bentin S, Mc Car thy G, Wood CC. Event-related po ten tials, lex i cal de ci sion and se man tic prim ing. Electroencephalogr Clin Neurophysiol 1985;60:343 55. 3. Federmeier KD, Kutas M. A rose by any other name: long term mem ory struc ture and sen tence pro cess ing. J Mem Lang 1999;41:469 95. 4. Mitch ell PF, An drews S, Fox AM, Catts SV, Ward PB, McConaghy N. Ac tive and pas sive at ten tion in schizo phre nia: an ERP study of in for ma tion pro cess ing in a lin guis tic task. Biol Psychol 1991;32:101 24. 5. Van Petten C, Kutas M, Kluender R, Mitchiner M, Melsaac H. Fractionating the word rep e ti tion ef fect with event-related po ten tials. J Cogn Neurosci 1991;3:131 50. 6. Gar cia-larrea L, Cézanne-Bert G. P3, pos i tive slow wave and work ing mem ory load: a study on the func tional cor re lates of slow wave ac tiv ity. Electroencephalogr Clin Neurophysiol 1998;108:260 73. 7. Halgren E. In sights from evoked po ten tials into the neuropsychological mech a - nisms of read ing. In: Scheibel AB, Wechs ler AF, ed i tors. Neurobiology of higher cog ni tive func tion. New York: Guilford Press; 1990. p 103 49. 8. Meyer DE, Schvanenveldt RW. Fa cil i ta tion in rec og niz ing pairs of words: ev i - dence of a de pend ence be tween re trieval op er a tions. J Exp Psychol 1971;90:227 34. 9. Condray R, Steinhauer SR, Co hen JD, Van Kammen DP, Kasparek A. Modulation of lan guage pro cess ing in schizo phre nia: ef fects of con text and haloperidol on the event re lated-potential. Biol Psy chi a try 1999;45:1336 55. 10. Koyama S, Nageishi Y, Shimokochi M, Hokama H, Miyazato Y, Miyatani M, and oth ers. The N400 com po nent of event-related po ten tials in schizo phrenic pa - tients: a pre lim i nary study. Electroencephalogr Clin Neurophysiol 1991;78:124 32. 11. Koyama S, Nageishi Y, Shimokochi M, Hokama H, Miyazato Y, Miyatani M, and oth ers. ERPs in schizo phrenic pa tients dur ing word rec og ni tion task and re - ac tion times. Electroencephalogr Clin Neurophysiol 1994;92:546 54. 12. Besche-Richard C, Passerieux C. Que nous apprennent les études de re con nais - sance de mots chez les schizophrènes? Ann Psychol 2001;101:521 43. 13. Ad ams J, Faux SF, Nestor PG, Shen ton M, Marcy B, Smith S, and oth ers. ERP ab nor mal i ties dur ing se man tic pro cess ing in schizo phre nia. Schizophr Res 1993;10:247 57. 14. Grillon C, Ameli R, Glazer WM. N400 and se man tic cate gori sa tion in schizo - phre nia. Biol Psy chi a try 1991;29:467 80. 15. Matsuoka H, Matsumoto K, Yamazaki H, Sakai H, Miwa S, Yoshida S, and oth - ers. Lack of rep e ti tion prim ing ef fect on vi sual event-related po ten tials in schizo - phre nia. Biol Psy chi a try 1999;46:137 40. 16. Nestor PG, Mat thew OK, O Donnell BF, Smith L, Niznikiewicz M, Shen ton M, and oth ers. Ab er rant se man tic ac ti va tion in schizo phre nia: a neurophysiological study. Am J Psy chi a try 1997;154:640 6. 17. Niznikiewicz MA, O Donnell BF, Nestor PG, Smith L, Law S, Karapelou M, and oth ers. ERP as sess ment of vi sual and au di tory lan guage pro cess ing in schizo phre nia. J Abnorm Psychol 1997;106:85 94. 18. Ohta K, Uchiyama M, Matsushima E, Toru M. An event-related po ten tial study in schizo phre nia us ing Jap a nese sen tences. Schizophr Res 1999;40:159 70. 19. Olichney JM, Iragui VJ, Kutas M, Nowacki R, Jeste DV. N400 ab nor mal i ties in late life schizo phre nia and re lated psy cho ses. Biol Psy chi a try 1997;42:13 23. 20. Salis bury DF, O Donnell BF, McCarley RW, Nestor PG, Shen ton ME. Event-related po ten tials elic ited dur ing a con text-free ho mo graph task in nor mal ver sus schizo phrenic sub jects. Psychophysiology 2000;37:456 63. 21. Salis bury DF, Shen ton ME, Nestor PG, McCarley RW. Se man tic bias, ho mo - graph com pre hen sion and event-related po ten tials in schizo phre nia. Clin Neurophysiol 2002;1:383 95. 22. Strandburg RJ, Marsh JT, Brown WS, Asarnow RF, Guthrie D, Harper R, and oth ers. Event-related po ten tial cor re lates of lin guis tic in for ma tion pro cess ing in schizo phren ics. Biol Psy chi a try 1997;42:596 608. 23. Passerieux C, Segui J, Besche C, Che va lier JF, Widlöcher D, Hardy-Baylé MC. Het er o ge ne ity in cog ni tive func tion ing of schizo phrenic pa tients eval u ated by a lex i cal de ci sion task. Psychol Med 1997;27:1295 302. 24. An drews S, Shel ley AM, Ward PB, Fox A, Catts SV, McConaghy N. Event-related po ten tials in di ces of se man tic pro cess ing in schizo phre nia. Biol Psy chi a try 1993;34:443 58. 25. Kay SR, Fiszbein A, Opler LA. The pos i tive and neg a tive syn drome scale (PANSS) for schizo phre nia. Schizophr Bull 1987;13:261 74. 26. Andreasen NC. Thought, lan guage and com mu ni ca tion dis or ders. I: Clin i cal as - sess ment, def i ni tion of terms and eval u a tion of their re li abil ity. Arch Gen Psy - chi a try 1979;36:1315 21. 27. Hardy-Baylé MC, Hardy P, Dantchev N. Stratégies et moyens thérapeutiques en psychiatrie. Paris: Doin; 1993. 28. Binois R, Pichot P. Test de vocabulaire. Paris: Edi tions du Cen tre de Psychologie Appliquée; 1959. 29. Silva-Pereyra J, Har mony T, Villanueva G, Fernandez T, Ro dri guez M, Galan L, and oth ers. N400 and lex i cal de ci sions: au to matic or con trolled pro cess ing? Clin Neurophysiol 1999;110:813 24. 30. McCarley RW, Niznikiewicz MA, Salis bury DF, Nestor PG, O Donnell BF, Hirayasu Y. Cog ni tive dys func tion in schizo phre nia: uni fy ing ba sic re search and clin i cal as pects. Eur Arch Psy chi a try Clin Neurosci 1999;249(Suppl 4):IV/69 IV/82. 31. Henik A, Nissimov E, Priel B, Umansky R. Ef fects of cog ni tive load on se man tic prim ing in pa tients with schizo phre nia. J Abnorm Psychol 1995;104:576 84. 32. Henik A, Priel B, Umansky R. At ten tion and automaticity in se man tic pro cess ing of schizo phrenic pa tients. Neuropsychiatry Neuropsychol Behav Neurol 1992;5:161 9. 33. Neely JH, Keefe DE. Se man tic con text ef fects on vi sual word pro cess ing: a hy - brid pro spec tive ret ro spec tive pro cess ing the ory. In: Bower JH, ed i tor. The psy - chol ogy of learn ing and mo ti va tion: ad vances in re search and the ory. New York: Ac a demic Press; 1989. p 207 48. 34. Neely JH. Se man tic prim ing ef fects in vi sual word rec og ni tion: a se lec tive re - view of cur rent find ings and the o ries. In: Besner D, Humphreys G, ed i tors. Ba sic pro cesses in read ing: vi sual word rec og ni tion. Hillsdale (NJ): Erlbaum; 1991. p 264 336. 35. Kutas M, Federmeier KD. Electrophysiology re veals se man tic mem ory use in lan guage com pre hen sion. Trends Cogn Sci 2000;4:463 70. Manu script re ceived Au gust 2002, re vised, and ac cepted De cem ber 2002. 1 Assistant Professor, Université Paris VIII, Equipe de Re cher che en Psychologie Clinique, Paris, France. 2 Hos pi tal Prac ti tio ner, Ser vice de Psychiatrie, Cen tre hospitalier de Ver - sailles, Ver sailles, France. 3 Assistant Pro fes sor, Université Paris VIII, Equipe de Re cher che en Psychologie Clinique, Paris, France. 4 Pro fes sor and Hos pi tal Prac ti tio ner, Ser vice de Psychiatrie, Cen tre hospitalier de Ver sailles, Ver sailles, France. Ad dress for cor re spon dence: Dr M Kostova, Equipe de Re cher che en Psychologie Clinique, Université Paris VIII, 2 rue de la Liberté, 93526 Saint-Denis Cedex 02, France e-mail: milena.kostova@univ-paris8.fr 622 Can J Psy chia try, Vol 48, No 9, October 2003

An Electrophysiologic Study: Can Seman tic Con text Pro cesses Be Mobi lized in Patients With Thought-Disordered Schizo phre nia? Résumé : Une étude électrophysiologique : peut-on mobiliser les processus du contexte sémantique chez les patients souffrant de schizophrénie avec trou ble de la pensée? Objectif : Cette étude visait à vérifier l hypothèse selon laquelle le renforcement de la struc ture du matériel ver bal pourrait apporter une amélioration des processus de con cep tuali sation chez les pa tients souffrant de schizophrénie avec trou ble de la pensée déclaré. Méthode : Un to tal de 38 pa tients schizophrènes avec trou ble de la pensée déclaré et 24 sujets témoins assortis en santé ont effectué 2 tâches de décision lexicale, comprenant 2 niveaux de struc ture contextuelle (avec 16,7 % et 33 % de mots reliés, respectivement). Les composantes N400 du potentiel lié aux événements, la composante pos i tive tardive (LPC) et les vari ables comportementales (temps de réaction et pourcentages d erreur) ont été analysées. Résultats : Un effet de structuration du contexte a été observé dans les LPC, mais pas les N400. Chez les sujets schizophrènes, les anom a lies des N400 (accroissement d amplitude pour les mots reliés et réduction de l effet N400) ont persisté dans les deux con di tions de structuration du contexte. De même, une réduction de l amplitude des LPC dans la catégorie de mots non reliés ainsi qu une réduction de l effet LPC ont été observées chez ces pa tients. Con clu sions : Les pa tients schizophrènes avec trou ble de la pensée déclaré n ont pas profité de la structuration du contexte afin de mettre en oeuvre des stratégies d intégration du contexte. Ce déficit sem ble être sta ble. Les résultats sont discutés dans le cadre du modèle publié antérieurement sur la compréhension du langage. Can J Psy chi a try, Vol 48, No 9, October 2003 623