The late positive component (P300) and information processing in sentences

Electroencephalography and Clinical Neurophysiology, 1975, 38:255-262 © Elsevier Scientific Publishing Company, Amsterdam - Printed in The Netherlands

255

T H E LATE POSITIVE C O M P O N E N T (P300) A N D I N F O R M A T I O N P R O C E S S I N G IN S E N T E N C E S 1 DAVID FRIEDMAN, RICHARD SIMSON, WALTER RITTER 2 AND ISABELLE RAPIN

Saul R. Korey Department of Neurology, Rose F. Kennedy Centerfor Research in Human Development nad Mental Retardation, Albert Einstein College of Medicine, Bronx, N.Y. 10461 (U.S.A.) (Accepted for publication: October 3, 1974)

Evoked potential correlates of the differential hemispheric processing of linguistic and nonlinguistic stimuli have been reported by many investigators for both visually-presented (e.g., Buchsbaum and Fedio 1969, 1970; Vella et al. 1972) and acoustically-presented (Cohn 1971; Morrell and Salamy 1971; Wood et al. 1971; Matsumiya et al. 1972) stimuli. The majority of these investigators have been concerned with the demonstration of hemispheric asymmetries in evoked potential amplitude as a correlate of the differential hemispheric processing of verbal and non-verbal information. While the issue of the cerebral localization of language function and whether it is reflected in the evoked potential is important, it has obscured the equally critical area of the evoked potential correlates of semantic, syntactic and phonologic information processing in language behavior. The "late positive component" or P300 (Sutton 1969), or the "Association Cortex Potential" (Ritter et al. 1972) is an ideal brain potential for studying linguistic processing, since it has been shown to be sensitive to the operation of cognitive variables (see Sutton 1969 and Karlin 1970, for reviews). Only a few reports of the presence of P300 in experimental paradigms of a linguistic nature have appeared. Donchin et al. (1973) reported significant left greater than right hemisphere asymmetries for both P300 and CNV 1 The research reported herein was supported in part by Grants NS 2503, NS 3356, 2T1 N 55325 from the National Institute of Neurological Diseases and Stroke, and MH 06723 from the National Institutes of Mental Health, U.S. Public Health Service. 2 Also at the Department of Psychology, Herbert H. Lehman College of the City University of New York.

potentials in a situation where the subject had to guess whether the letter A or B would appear as $2. When the subject was required to make a differential response to the letter A or B, P300 and CNV were larger on the right. In a report by Marsh and Thompson (1973), $1, a flash, preceded a verbal stimulus on some blocks, and preceded a slanted line on other blocks. Averaging the blocks separately, neither CNV, nor P300 to $1 were found to be asymmetric. Both of these reports utilized the classical CNV paradigm, and, other than looking for the presence of asymmetries, these findings did not bear on the issue of the evoked potential correlates of linguistic processing. Shelburne (1972, 1973) utilized the "information delivery" paradigm of Sutton et al. (1967), where the third letter of a sequence of letters either completed a nonsense syllable or a real word. P300 waves were present to the critical third letter, but not to the other two, and were of similar amplitude whether the word was real or nonsense, thus demonstrating the information delivery effect, but not uniquely reflecting anything related to language. In addition, the P300 waves seen were not asymmetric. Since the majority of language phenomena take place in the form of sentences (either on the printed page or in connected discourse), we felt that the use of words comprising a sentence might be a better way of obtaining evoked potential correlates related to linguistic processing. A frequent phenomenon in language is that the understanding of words early in a sentence is determined by words produced later in the sentence (Warren and Warren 1970; Slobin 1971). We

D. FRIEDMAN et al.

256

employed a situation in which the understanding of key words in a sentence was determined either late (condition 1) or early (condition 2) within the sentence. Since our subjects were involved in decision processes, we expected the presence of late positive components to stimuli which delivered sentence information to the subject. METHOD

Eight right-handed (4 male and 4 female, naive with respect to evoked potential experiments) college students between the ages o f 19 and 23 served in the experiment and were paid for their participation. The subject was comfortably seated in a sound-damped, electrostatically shielded room facing a constantly illuminated white ptexiglass screen (202 ft.-candles). Stimuli were black on white 35 mm transparencies rear-projected on to the screen at a rate of one every 2.4 sec by a Carousel projector (Ectagraphic Model B)with an attached shutter (Polymetrics Company) which allowed a 40 msec presentation time for each slide. The stimuli were approximately 220 ft.-candles in intensity .A small, black dot in the center of the screen served as a fixation point. The visual angle subtended by the stimuli varied from 2 deg t o 3 deg 42 min depending on the number of tetters in the word, and was well within foveal vision. The timing sequence for a single trial can be seen in Fig. I. The start of the trial was signalled by a slide with two centered dots followed by the sequential presentation of the six words com-

WORDS

SLIDES

the

I-eel) heel

is

on

the

shoe

~2.4 SEC:-N

I

I

I

I

I

I

n

n

n

H

n

blank

n 4 0 ms

EEODATA lOOms

prising the sentence, and ending with a slide with two centered dots. Three sentences were used throughout the experiment: The wheel is on the axle." The heel is on the shoe. and The peel is on the orange. Subjects were told the form of the sentences in advance. The subject's task was to report verbally the second word during the 2.4 sec between trials. In condition 1. the first grapheme of the second word was omitted (-eel). The subject. therefore, could not know the meaning of the second word uritil the last word was presented. In condition 2. the second word was presented entirely, so that it was not necessary to wait until the end of the sentence to know its meaning. In fact. in condition 2 the subject could infer the last word after seeing the second word. In this way, we were able to study the evoked potentials to words which did and did not convey the semantic information necessary for completion of the task. A run consisted of the presentation of 20 sentences (with a short break following the tenth), wherein each of the three sentences occurred approximately the same number of times (6, 7.7) in a random order. There were four runs of each condition, with the order of conditions counterbalanced across runs. Electrodes were placed over the right eye (OD), at Cz, and at left and right temporo-parietal sites located midway between Pz and the mastoid process. The reference electrode was on the tip of the nose. Electrical activity was amplified by Tektronix type 2A61 differential amplifiers with

n SLIDE ON TIME~--~

I--I-71 BASELINE "~-~l ~ "

I--1----1 7T--1

V7--I

I SEC ANALYsIsRESPONSE

STIMULUS Fig. 1. Timing sequence for the presentation of a single sentence.

VT-N

blank

257

THE LATE POSITIVE COMPONENT IN SENTENCES

CONDITION

1 axle

I.

THE

-eeJ

'22o / C~4"~

IS

ON

-

THE

A_

NI40

~"~'/

~'~'~l

",,/

4-

CONDITION 2

THE

hoel whool

IS

ON

THE

axle L

• .,"

~j

t,,,,,,,,,,

'"'-- -..,j i , / ' "~--

~..,j

,,.

T,,,,,,,,,,

_ j\._,,..~_..

._../"

t,,,,,,,,,,

1100 MSEC

Fig. 2. G r a n d Mean evoked responses averaged across all 8 subjects for each word in each condition at electrode Cz.

low and high band passes of 0.06 and 50 c/see at a gain of 20,000. The EEG and pulses corresponding to stimulus onset were stored on FM magnetictape(MnemetronTapeDeck),andaveraged off-line by a LINC 8 computer and written out on a Cal-Comp X-Y plotter. The basic writeout consisted of the evoked response to each of the six words in the sentence averaged separately for each condition across runs, yielding an N of 80 per averaged response. In addition, to monitor possible CNV effects, longer analysis times encompassing the first and second or the fifth andsixthwordswereexaminedinseveralsubjects, Amplitude measurements were made from the 100 msec pre-stimulus baseline to negative and positive peaks at each electrode site for each word in each condition. Peak latencies from stimulus onset were rounded off to the nearest 10 msec. 2 by 6 analyses of variance for repeated measures (Winer 1962) were performed on the amplitude and latency of each component at each electrode site to determine if condition, word position, and their interaction had significant effects,

across all 8 subjects to each of the words in the sentence from each condition at electrode Cz are presented in Fig. 2. The typical response consisted of a negative deflection (N,: m e a n = 140 msec), followed by a positive deflection (P2 : mean = 220 msec), another negative deflection (N2: mean = 280 msec), and a final positive deflection (P300), which ranged from 300 to 600 msec in latency depending on the experimental condition. While there was variability among subjects, within a given subject, the amplitude and latency of the various components to each of the six words in the two conditions were essentially similar, with the exception of P300 as described below. Since other investigators (e.g., Vella et al. 1972; Donchin et al. 1973) have reported evoked potential amplitude asymmetries between the left and right hemispheres to visually presented language and non-language stimuli, all components were tested for significant asymmetries using the t-test for correlated means. No significant asymmetries were found. This is consistent with the findings of Friedman et al. (1975) for auditory language stimuli, and Shelburne (1972, 1973) for visual language stimuli.

RESULTS

Amplitude

The Grand Mean evoked responses averaged

effects The mean baseline to N1 amplitudes for each

D. FRIEDMAN et al.

258

I

I

I

I

I

I

I

I

f

hemisphere. However, if N z enhancement reflected differential attention to the task-related word of a given condition, then these same effects should have occurred to the sixth word of condition 1, but in fact did not. Indeed, word position assessed by tests for simple effects had no effect on N1 amplitude for condition 1 at any of the electrode sites. For components P2 and N 2 no consistent pattern or significant relationships were found. Fig. 4 presents the mean baseline to P300 amplitudes for each of the words in each condition. At each electrode location, word position significantly affected P300 amplitude (P < 0.01 ). Tests for simple effects of word position for each condition showed significant variation for both conditions l and 2 at all electrode sites (P<0.01).

LEFT HEMISPHllE

z

:£ <

I

I

I

z6

20-.....

8

|tGHT HEMISPHERE

CONDt

/

COND2

vl|tlx

15-6

i •\

10--

4

t -

1/ •

2-

t t

t

tt

5-

i THE

I HEEL

I IS

I ON

I THE

f I

SHOE

T

I

I

(.EEL)

Fig. 3. Mean baseline to Peak N] amplitude at all electrode sites for all words in conditions 1 and 2.

of the words in each condition are presented in Fig. 3. Word position had a significant effect on N1 amplitude (P<0.05) at Cz and the left hemisphere electrode site, but tests for simple effects showed that this occurred only for condition 2 (P < 0.05). Further tests for simple effects of condition at each word position showed that N1 amplitude to the second word of condition 2 (wheel, heel,peel, which delivered the information required for the task) was significantly (P <0.05) larger than N 1 amplitude to the second word of condition 1 (-eel, which did not deliver information), but only at the left hemisphere electrode. It is possible that this effect was a reflection of differential attention (Hillyard et al. 1973) to the word the subject knew would deliver the taskrelated information. It is intriguing that the difference in N 1 amplitude to the second words between conditions occurred only over the left

20--

tE~t ~eMIseHeee

i

15-

z

10-

5-

<

I

~--T-

I

]

I

O

o L

.....

20O

¢OND

1

COND

2

lS-

/

/

//

lO-

S-

I THE

! HEEL (-EELI

1 IS

I ON

I

I

THE

SHOE

Fig. 4. Mean baseline to peak P300 amplitudes at all electrode sites for all words in conditions l and 2.

259

THE LATE POSITIVE COMPONENT IN SENTENCES

FL

IK

GJ

I',..,'

~, ....

R

3

i

CZ

t

-I"

/s

-

..... COND1

1100 MSEC.

--

COND2

Fig. 5. Comparisons of P300 amplitude for 3 subjects between the last words of each condition.

Multiple comparisons showed that in both conditions the last word produced significantlylarger P300 amplitude (P < 0.05) than each of the other words in the sentence, whether it provided the subject with task-related information (condition 1) or provided no further information (condition 2). In addition, tests for simple effects of condition at each word position showed that P300 amplitude to the last word of condition 1 was significantly larger (P<0.05) than P300 amplitude to the last word of condition 2 at the left and right hemisphere sites, but not at Cz. This is shown for three subjects in Fig. 5.

Latency findings Significant latency effects were restricted to the P300 component. Mean P300 latency to each word in each condition at all electrode sites is presented in Fig. 6. At all electrode sites word position had a significant effect on P300 latency (P<0.01). In addition, at Cz and the left hemisphere electrode, the condition by word

position interaction was significant (P<0.05). This interaction was due to the fact that P300 latencies to second and sixth position words which delivered information were longer than second and sixth position words which did not deliver information. Since the words which delivered information were in different positions depending upon the condition (second word for condition 2, sixth word for condition 1), this latency reversal effect accounted for the interaction. These comparisons for second and sixth words can be seen in Fig. 6. Tests for simple effects of condition at each word position, showed that all these comparisons were significant (P<0.05). In other words, when information was delivered via the second word (condition 2), latency of the P300 component was longer to that word than to -eelofcondition 1, while the reverse relationship held when the last word delivered the information - - P300 latency to the last word of condition 1 was longer than P300 latency to the last word of condition 2.

260

D. FRIEDMANet al.

i

5IX}- i

revealed the presence of CNVs in some circumstances but not in others. The effects on P300, however, were observed whether or not a CNV was present, and regardless of its magnitude.

IIIGHT R~*MISPH|I|

I

I)ISCUSSION 3OO

I

o=~ soo] _z

I

I

I

I~----T-'--

[

I

LEFT N t MlSPHltE

1.2

o a.

3OO Z

I

I

I

q--

,<

,o. 7

300

w RTIEX

v

l

!

I

THE

HEEL

IS

ON

......

¢OND|

- -

COND2

I

t

THE

SHOE

C_EEL]

Fig. 6. M e a n P300 latency at all electrode sites for all w o r d s in c o n d i t i o n s 1 and 2.

Evaluation of the significant word position effect by tests for simple effects showed that at Cz and the left hemisphere site this effect was significant for both conditions (P<0.01), but only for condition 2 at the right hemisphere site (P<0.01). Multiple comparisons revealed that the word which delivered task-related information, whether it was in the second (condition 2) or sixth (condition 1) position, always had a longer P300 latency than all the other words in the sentence (P<0.05). This can be clearly seen in Fig. 6. C N V findinys While there are reports in the literature indicating a possible relationship between the termination o f a C N V and the latency of P300 (e.g., Donald 1968), examination of the activity between the first and the second and the fifth and the sixth words of the sentence in both conditions

C o n t r a u to tile usual linding (e.q., Sutton c't al. 1965: Tueting et al. 1971; Friedman et al. t973) information delivery had its strongest effects on P300 latency, and not on P300 amplitude. There was a P300 amplitude difference favoring the last word of condition 1 (which delivered information) over the last word of condition 2 (which did not deliver information). However, there was no difference in P300 amplitude between the second word of condition 2, where it delivered information, and the second word of condition 10 where it did not. Furthermore, in condition 2, P300 amplitude to the second word (which delivered informatmn) was not significantly different from P300 amplitude to the first, third, fourth and fifth words, all of which were redundant (see Fig. 3). Moreover, whether or not the last word delivered information, P300 amplitude to the last word in both conditions 1 and 2 was significantly larger than P300 amplitude to any of the other words in the sentence. Even in condition 2. where the second word did deliver information, P300 amplitude to the last word was greater than P300 amplitude to the second word. However. whenever a word delivered information. P300 latency was significantly longer to that word than to any other word in the sentence and. in addition, was significantly longer than the same position word in the other condition (which did not deliver information). Comparisons made between conditions to words which were redundant (the. ts, on) showed no significant P300 latency differences These findings led us to the following conclusions: In language experiments, uncertainty has its greatest effect on P300 latency and not amplitude (i.e.. the latency reversal effect between key words in conditions 1 and 2). The evoked potential to the word which ends the syntactic structure produces a larger P300 than all the other words within that structure. This we call "'syntactic closure". Presumably, the brain processes language stimuli in terms of their meaning

261

THE LATE POSITIVE COMPONENT IN SENTENCES

and syntactic structure simultaneously. These are independent matters, in that syntactic structure need not have meaning, and humans are able to determine if word sequences have syntactic structure whether or not they have meaning. When a syntactic structure, such as the sentence, ends, the brain presumably recognizes it as a language unit. The enhancement of P300 to the last word of the sentence regardless of meaning and the point of task-related information delivery within the sentence, suggests that this enhancement of P300 is a correlate of"syntactic closure". The finding of no significant asymmetries in any of the components studied supports previous findings of Friedman et al. (1975). An extended discussion of evoked potential asymmetry as a correlate of linguistic processing can be found in that paper. An unusual finding, which is not explained by any of the theoretical formulations concerning the genesis of P300 (e.9., Sutton 1969; Squires et al. 1973a, b) was the presence of P300s to all the words in the sentence regardless of whether or not they delivered information to the subject. Both Tueting et al. (1971) and Friedman et al. (1973) found small amplitude P300 components in a situation where stimulus events were completely predictable. Recently, Donchin et al. (in press), using an RT paradigm, reported the presence of large amplitude P300 waves to predictable letter stimuli (A or B). Friedman et al. (1975) also found small amplitude P300 components to acoustically presented word and sound stimuli when the subject simply listened attentively to the stimuli. The presence of P300 waves to completely predictable word stimuli in the present experiment, supports the speculation made in our earlier paper (Friedman et al., 1975) that the P300 system is engaged when language stimuli are presented and the subject has a task. SUMMARY

Averaged visual evoked potentials to sequentially flashed words comprising a sentence were recorded from vertex and left and right temporoparietal electrodes in 8 right-handed subjects. In condition 1 the sentence took the form: The -eel is on the shoe, in which the first grapheme was

omitted from the second word, so that the subject did not know the meaning of the second word until he viewed the last word. In condition 2, the sentence took the form : The heel is on the shoe, in which the second word was given and the last word provided no further information. P300 latency to words which delivered information (last word of condition 1, second word of condition 2) were significantly longer than P300 latency to any of the other words in the sentence, as well as to the same position word in the other condition. Comparisons of P300 latencies to redundant words (the, is, on) within and between conditions showed no significant differences. P300 amplitude to the last word was significantly larger than P300 amplitude to any of the other words within the sentence, even in condition 2 where the second word delivered information. The major effect of information delivery was on P300 latency, while "syntactic closure" had its major effect on P300 amplitude. The fact that evoked potentials to all words had P300 components was attributed to the engagement of the P300 system whenever task-related language stimuli are used. RESUME LA COMPOSANTE POSITIVE TARDIVE ( P 3 0 0 ) ET LE PROCESSUS

D'INFORMATION

A

L'INTERIEUR

DE

PHRASES

Les potentiels 6voqu6s visuels moyens h des mots 6clair6s s6quentiellement compris dans une phrase ont 6t6 enregistr6s au niveau du vertex et des 61ectrodes temporopari6tales droite et gauche chez 8 sujets droitiers. En condition l, la phrase prend la forme The -eel is on the shoe dans laquelle le premier graph6me du second mot est supprim6 de telle sorte que le sujet ne connait pas la signification du second mot jusqu'h ce qu'il Pit vu ie dernier mot. En condition 2, la phrase prend la forme The heel is on the shoe dans laquelle le second mot est donn6 et le dernier mot ne fournit aucune information suppl6mentaire. La latence du P300 ~ des roots qui donnent l'information (le dernier mot en condition l, le second mot en condition 2) est significativement plus longue que la latence du P300 5. n'importe lequel des autres mots dans la phrase, de m~me que la latence aux mots de m~me position dans l'autre condition. La comparaison des latences de P300

262

/i des mots redondants (the, is, on)/l l'int6rieur d'une m~me condition et d'une condition/t l'autre ne montre aucune diff6rence significative. L'amplitude du P300 au dernier mot est de fa~on significative plus grande que l'amplitude du P300 /l n'importe lequel des autres mots/t l'int6rieur de la phrase, m~me en condition 2 off le second mot fournit l,information. L'effet majeur du fait qu'une information est fournie s'exerce sur la latence du P300, tandis que la "fermeture syntaxique" a son effet majeur sur l'amplitude du P300. Le fait que les potentiels 6voqu6s ~t tous les mots ont des composantes de P300 est attribu6 l'engagement du syst+me des P300 chaque lois que sont utilis6s des stimuli de langage li6s "~ la tfiche. The authors would like to thank Dr. Herbert G. Vaughan Jr., for the use of the LINC 8 computer and Ms. Julie Wong, for computer programs used in this study. Drs. Samuel Sutton and Patricia Tueting read an earlier version of this report and made helpful suggestions.

REFERENCES BUCHSBAUM,M. and FEDIO,P. Visual information and evoked responses from the left and right hemispheres. Electroeneeph, clin. Neurophysiol., 1969, 26: 266-272. BUCHSBAUM, M. and FEDIO, P. Hemispheric differences in evoked potentials to verbal and non-verbal stimuli in the left and right visual fields. Physiol. Behav., 1970, 5: 207-210. COHN, R. Differential cerebral processing of noise and verbal stimuli. Science, 1971, 172: 599-601. DONALD, i . W. Electrocortical correlates of fixedJbreperiod decision tasks. Doctoral Dissertation, McGill University, 1968. DONCHIN, E., JOHNSONJR., R., HERNONG,R. and KUTAS,M. Covariation of the magnitude of the CNV and PD00 as a function of the subject's task. Paper read at the llI int. Congr. event-related slow potentials of the brain. Bristol, 1973. DONCHIN, E., KUBOVY, M., KUTAS, M., JOHNSON, R. and HERRING, R. Graded changes in evoked response (P300) amplitudes as a function of cognitive activity. Electroenceph, clin. Neurophysiol., in press. FRIEDMAN, D., HAKEREM,G., SUTTON, S. and FLEISS,J. L. Effect of stimulus uncertainty on the pupillary dilation response and the vertex evoked potential. Electroenceph. clin. Neurophysiol., 1973, 34: 475~84. FRIEDMAN,D., SIMSON,R., RITTER, W. and RAPIN, I. Cortical evoked potentials elicited by real speech words and human sounds. Electroenceph. clin. Neurol,." ' 1975, 38: 13-19. HILLYARD, S. A., HINK, R. F., SCHWENT,V. L. and PICTON, T. W. Electrical signs of selective attention in the human

D. FRIEDMAN et al.

brain. Science, 1973, 182:177 180.

KARLIN, L. Cognition, preparation and sensory evoked potentials. Psychol. Bull., 1970, 73: 122-136. MARSH, G. R. and THOMPSON, L. W. Effect of verbal and non-verbal set on hemispheric asymmetries in the CNV. In W. C. MCCALLUM and J. R. KNOTt (Eds.), Eventrelated slow potentials o! the brain: Their relations to behavior. Electroenceph. clin. Neurophysiol., 1973 Suppl. 33: 195-200. MATSUMIYA, Y., TAGLIASCO, V., LOMBROSO, C. T. and GOODGLASS, H. Auditory evoked response : Meaningfulness of stimuli and interhemispheric asymmetry. Science, 1972, 175 : 790 792. MORRELL, L. and SALAMY,J. Hemispheric asymmetry of electrocortical responses to speech stimuli. Science, 1971, 174:164~166. RITTER, W., SIMSON,R. and VAUGHANJR., H. G. Association cortex potentials and reaction time in auditory discrimination. Electroenceph. olin. Neurophysiol., 1972, 33:547-555. SHV.LnURNEJR., S. A. Visual evoked responses to word and nonsense syllable stimuli. Electroenceph. clin. Neurophysiol., 1972, 32:17 25. SHELBURNEJR., S. A. Visual evoked responses to language stimuli in normal children. Electroenceph. clin. Neurophysiol., 1973, 34: 135-143. SLOBIN, D. I. Psyeholinouistics. Scott, Foresman, Glenview. 1971. SQUIRES,K., HILEYARD,S. A. and LINDSAY,P. Cortical potentials evoked by confirming and disconfirming feedback following an auditory discrimination. Percept. Psychophys., 1973a, 13:25 3I. SQUIRES,K. C., HILLYARD, S. A. and LINDSAY,P. H. Vertex potentials evoked during auditory signal detection : Relation to decision criteria. Percept. Psychophys., 1973b, 14:265 272. SUTTON, S. The specification of psychological variables in an averaged evoked potential experiment. In E. DONCHIN and D. B. LINDSLEY (Eds.), Average evoked potentials: Methods, results and evaluations. NASA, SP-191, Washington, D.C, 1969: 237--262. SUTTON, S., BRAREN,M., ZUBIN,J. and JOHN, E. R. Evoked potential correlates of stimulus uncertainty. Science, 1965, 150:1187-1188. SUTTON, S., TUETING, P., ZUB1N,J. and JOHN, E. R. Information delivery and the sensory evoked potential. Science, 1967, 155: 1436-1439. TUETING, P., SUTTON, S. and ZUmN, J. Quantitative evoked potential correlates of the probability of events. Psychophysiolooy, 1971, 7: 385-394. VAUGHANJR.; H. G. and RITTER, W. The sources of auditory evoked responses recorded from the scalp. Electroenceph. c#n. Neurophysiol., 1970, 28: 360-367. VELLA, E. J., BUTLER, S. A. and GLASS, A. Electrical correlates of right hemisphere function. Nature (Lurid.), 1972, 236: 125-126. WARREN, R. M. and WARREN, R. P. Auditory illusions and confusions. Sci. Amer., 1970, 223:30 36. WINER, B. J. Statistical principles in experimental desion. McGraw Hilt, New York, 1962. WOOD, C. C., GOFF, W. R. and DAY, R. S. Auditory evoked potentials durin~ speech perception. Science, 1971, 173: 1248-125i.