- Email: [email protected]
A waveform dissimilarity of auditory evoked potentials induced by the stimulation of meaningful nouns
133
Biological Psychology 18 (1984) 133-147 North-Holland
A WAVEFORM POTENTIALS NOUNS Hideaki
NINOMIYA
Department Accepted
DISSIMILARITY OF AUDITORY EVOKED INDUCED BY THE STIMULATION OF MEANINGFUL
of Psychiatry, for publication
* and Terutika
IKEDA
**
Miyaraki
Medical College, Mlyaraki,
31 August
1983
Japan
Hemispheric asymmetry in language processing was investigated using an auditory evoked potential as the index for neurophysiological activity. Subjects were given the task of recognizing and classifying typical Japanese nouns, responding to nonsense syllables and responding to tone pips. Fourteen subjects participated and the AEPs of 12 subjects from the frontal, central and parietal regions were adopted for subsequent analyses. Correlation coefficients between the right/left AEPs from homologous scalp regions and a waveform similarity curve analysis that can indicate the similarity of two AEPs segmentally and successively, were adopted to provide the intra-individual symmetry index. Results were as follows: (1) The waveforms between the right/left AEPs induced by meaningful nouns were significantly dissimilar than those induced by nonsense syllables and tone pips; (2) the dissimilarities occurred mainly at 160 msec and 350 msec from the onset of the first syllable on the frontal and central regions. These findings may reflect the hemispheric asymmetry of neurophysiological activity in language processing.
1. Introduction The assumption that each hemisphere of the brain has different functions, and that the left hemisphere plays the dominant role in language processing was suggested by early clinical, anatomical and psychophysiological findings (e.g. Broca, 1861; Wernicke, 1874; Wada and Rasmussen, 1960; Kimura, 1967; Geschwind and Levitsky, 1968; Geschwind, 1970). In the neurophysiological field, to test the asymmetrical functions of the left/right hemispheres in language processing, many recent studies have been using the evoked potentials as an index. * Address Medical ** We wish Maxwell
requests for reprints to: Hide&i Ninomiya, Department of Psychiatry, Miyazaki College, 5200 Kihara, Kiyotake-cho, Miyazaki-gun, Miyazaki, Japan. to express our thanks to Ms. Y. Ibusuki and J. Ono for their assistance, to Mrs. G. and Mr. D. Nakahara for their helpful discussion.
0301-0511/84/$3.00
0 1984, Elsevier Science Publishers
B.V. (North-Holland)
134
H. NI~OWIIJYIand T. Ikeda / A waveform dissimilarity
of AEP b.v the me~n~nRful noun.s
Buchsbaum and Fedio (1970) first reported that the waveforms of visual evoked potentials that were induced by verbal and non-verbal stimuli were more dissimilar in the left hemisphere, as compared with the right, suggesting that hemispheric differences in the averaged evoked potentials were consistent with the hypothesis that each hemisphere of the brain had different roles. Investigations using auditory evoked poentials (AEPs) reported the same in the early 1970’s. Morrell and Salamy (1971) and Matsumiya, Tagliasco. Lombroso and Goodglass (1972) reported that the peak amplitude of Nl(Morrel1 and Salamy, 1971) or ‘ W’ wave (Matsumiya et al., 1972) in the left temporo-parietal area was larger than those in the right. Wood and Goff (1971) divided the AEP into two parts, i.e. the preresponse interval and the response interval, reporting that the amplitude of each sampling point of a task requiring the analysis of linguistic information was significantly different in the preresponse interval from those of a task which did not require the analysis of linguistic information. On the other hand, some studies failed to confirm the hypothesis of the hemispheric asymmetry (Friedman, Simson, Ritter and Rapin. 1975; Galambos, Benson, Smith, Schulman-Galambos and Osier, 1975; Grabow, Aronson, Rose and Green, 1980). Friedman et al. (1975) closely examined the experimental design and the statistical analysis of early studies, and concluded that the data demonstrating the hemispheric asymmetry of evoked potentials were far from convincing. Since Grabow et al. (1980) who replicated the study of Morrell and Salamy (1971) were unable to find the hemispheric differences, it could be suggested that the use of a simple measurement of the peak amplitudes was inadequate as the test for asymmetry. Brown, Marsh and Smith (1973. 1976) also used a correlational method similar to that of Buchsbaum and Fedio (1970). They compared the AEPs induced by various meaningful words and found that the words which were similar in sound but not in meaning, produced more dissimilar waveforms in the left hemisphere than in the right. They also suggested in their 1976 paper that their method of using the coefficient of correlation calculated over the entire evoked potential did not provide any indication of response epochs or components. Donchin (1979) suggested that since many cognitive variables might affect evoked potentials, it might be more appropriate to use multivariate methods of analysis. Molfese, Papanicolaou, Hess and Molfese (1979), using a principal component analysis, reported that on three of the factors, there was a significant effect as a function of meaning and that in each case the effect was noted over a short period. In most studies mentioned above, the compared evoked potentials were recorded successively at different times. Roth, Kopell, Tinklenberg, Huntsberger and Kraemer (1975) reported that investigating the test/re-test reliability of the averaged evoked potentials of the same person at successive times, the median correlation coefficient was formed to be 0.59. In other words, there
H. Ninomtya
and T. Ikeda / A waveform dissimilarrty of AEP by the meaningful
mum
135
was a moderate amount of intra-individual consistency. It is well known that the inter-individual variability of evoked potentials is large. One of the methods of eliminating the variability in evoked potentials caused by changes over some time, or differences between individuals, is to calculate a coefficient based on the data from the same individual on the basis of evoked potentials recorded simultaneously. Harmony, Ricard, Ostero, Fernandez, Llorente and Valdes (1973) reported that the evoked potentials of bilateral homologous ‘regions which were recorded simultaneously showed quite similar waveforms. For this reason coefficients calculated in this way would provide the index for symmetry. Perhaps the meaningful words would be sensed in both hemispheres but the recognition or classification of them would be processed mainly in the left hemisphere. Thus, it can be hypothesized that the synchronization of the left and right evoked potentials elicited by the meaningful words would be less than that elicited by pure tones. In the present study, the subjects were given the tasks of recognizing and classifying some typical Japanese nouns in two syllables, as well as responding to nonsense syllables and to tone pips, and a correlational analysis was adopted in two ways: (1) Correlation coefficients have been calculated on the basis of evoked potentials, recorded simultaneously, of the left and right homologous regions to provide the intra-individual symmetry index; (2) as Molfese et al. (1979) reported that the effect of meaning occurred at short periods, there was a possibility that the left and right desynchronization presented only for a short time during the processing of a meaningful stimulus. If such was the case, the analysis of the whole sweep time was unlikely to detect asymmetries. Therefore, the data were also subjected to an analysis in segments by a waveform similarity curve method that is a modified correlational analysis (Nakahara and Ikeda, 1981).
2. Methods 2.1. Subjects and setting Fourteen paid subjects (two females and twelve males, age 23-26) participated in the experiment. All were right handed and audiologically normal. They were seated on a reclining chair, located 1 m in front of a loud speaker in a sound-damped and electrostatically shielded room which was dimly lit (40 lux). Ag/AgCl disc electrodes were placed bilaterally on the scalp at the frontal, central and parietal regions according to the ten-twenty system. The linked-ear electrodes were used for a reference. EEGs were amplified by Nihon Koden AB-620G bioelectric amplifier (0.3 set in time constant and 100 Hz high-cut filter), and were recorded on a Sony UFR-70760 data recorder. The arousal
136
H. Ninomiya
and T. lkeda / A waveform dissmilarit~y of AEP by the meaningful nouns
level and eye movements were monitored by an EEG paper recorder. Two out of 14 records were omitted from subsequent analysis because of an excessive contamination by eye movements. 2.2. Stimulus The real voice of a female volunteer and tone pips were used for the stimuli, and two kinds of speech were used; one was meaningful words in two syllables, and the other was nonsense words in two syllables (PIPI, PEPE). An example of each stimulus is shown in fig. 1. Each verbal stimulus lasted for about 200 msec and the intensity of all the stimuli was adjusted to be as similar as possible. The mean onset of the second syllable of the meaningful words was 113 + 43 msec after the onset of the first syllable. As for the tone pips, two kinds of stimuli were used; one was 300 Hz, 65 dB SPL, 200 msec duration, and another was 600 Hz, 65 dB SPL, 200 msec duration. 2.3. Procedure There were three conditions: (a) Meaningful words condition (MW); (b) nonsense words condition (NW); (c) tone pips conditions (TP). Each condition consisted of 20 blocks of 5 successive stimuli. Examples of each block are shown in fig. 2. In the meaningful words condition, the words in two syllables (consonantvowel-consonant-vowel) were used as stimuli. In 20 blocks, there were two types of blocks, one consisted of 5 words from the same class, and the other consisted of 4 words from the same class mixed with the word from the
ZE 100 msec
300 Hz TONE PIPS Fig. 1. An example of stimuli in three conditions. Each stimulus is put into Sanei Signal Processor and drawn out by the Watanabe (window).
X-Y
recorder. The meaningful
word shown in the fig. is kaze
H. Ninomiya
(,)
and T. Ike-da / A waue/orm dissimilarity
k mo
(A\
nami
ki
of AEP
by the meaningful nouns
137
button press
I
nr (B)
(III)
(A)
pipi
I
pipi
pi i
300
300
300
i
epe
300
r
r
‘7sec
f
Fig. 2. An example of the blocks where a button press is required and where not required. (I) meaningful words condition, (II) nonsense words condition, (III) tone pips condition. (I) (A) all nouns are the class of natural phenomena, (B) mixed with plants (taka is bamboo). Meaningful nouns shown in English are kaze (window), kiri (fog), kumo (cloud), yuki (snow) and nami (wave). t-- : Beginning of the next block.
different class. The classes used were as follows: The natural phenomena; parts of body; personal ornaments; animals; fishes; birds; and plants. To check that the stimuli had been attented to and comprehended, the subjects were instructed to push a button at the end of each block if they thought the block had consisted of 5 words from the same class. In the nonsense words condition, there were the block of 5 PIPIs and that of 4 PIPIs with one PEPE and in the tone pips condition, the block of five 300 Hz tone pips and that of four 300 Hz with one 600 Hz. They were also instructed to push the button if 5 successive stimuli (i.e. one block) were identical. As the finger lifting or button press occurred only after a block of stimuli had been presented, the motor potential that contaminated the late component of auditory evoked potentials could be ignored. The nonsense words condition was used because a possibility existed that a human voice even without meaning might induce an asymmetry in the evoked potentials (Morrell and Salamy, 1971). Six subjects participated in the experimental sequence of (a)-(b)-(c) and other 6 subjects in (c)-(b)-(a) to counterbalance for order effect. 2.4. Data analysis The EEG of each subject was averaged over 100 times by Sanei Signal Processor 7T07A from a trigger pulse to 510 msec with a sampling rate of 1 point per 2 msec. The averaged potentials were written out as an analogue record by the Watanabe X-Y recorder, and also were punched out in a digital form by Kasio model 501 typuter. Punch tapes were processed by Canon BX-1 microcomputer for further analysis.
(a) Component analysis The NlLP2 peak-to-peak amplitude were measured, and R-values were calculated by the same method as that of Matsumiya et al. (1972). A R-value greater than 0.5 indicated that the left evoked potential was greater than the right, and smaller than 0.5, the right greater than the left. In the present paper, the Nl-P2 peak-to-peak amplitude was used to calculate the R-value. (b) Correlational analysis (i) Whole sweep time (510 msec): Pearson product moment correlation coefficients of evoked potentials from the left and right homologous scalp regions were calculated by Canon BX-1 on the basis of digitised punch-tapes. The calculated values indicated the degree of resemblance between the waveforms of the left and right evoked potentials. (ii) Waveform similarity curve: Essentially this method was a correlational technique. There were two methodological advantages, one was that the optimal sampling length (window) to calculate a correlation coefficient could be taken out, and another was that the window moved from the trigger point to the end of sweep time step by step on the sampling time. Therefore, it could indicate the degree of the resemblance between two evoked potentials segmentally and successively. Details of this method have been fully described in a previous paper (Nakahara and Ikeda, 1981). In the present paper, the sampling length of the waveform similarity curve was 118 msec (sampling point = 60), therefore correlation coefficients of two evoked potentials were calculated from O-118 msec to 3922510 msec successively (the total number of correlation coefficients was 197). The calculated value was dotted at the center of each sampling length (59, 61, 63, . . ., 451 msec).
3. Results As the mean percents of correct responses were not different among the three conditions (> 99%) it was suggested that each subject understood the task and attended to the stimuli. The grand averages of auditory evoked potentials of 12 subjects for each region at each condition are shown in fig. 3. Comparing Nl and P2 amplitudes among the three conditions, the Nl amplitudes of nonsense words and tone pips conditions were larger than those of the meaningful words condition and there was not much difference in the P2 amplitudes among the three conditions. 3. I. Component analysis The mean and s.d. of Nl-P2 amplitudes are shown in fig. 4. All the NlLP2 amplitudes of MW were significantly lower than those of NW and TP both in
: :Ij
H. Nlnomiyu
:
‘...,.
..
and T. Ikedcr / A waveform
dissimilarity
o/A EP by the meaningful
nouns
139
140
H. Ninomiya
and T. Ikeda / A waueform dissimilarity
FRONTAL
of A EP by the memingful
CENTRAL
nouns
PARIETAL
4v
10
10
10
: ::
:* :;
5
0
II
.
II
r-m
NW
TP
:
I
:+ :I 5
5
0
_
i rlw
TP
.
.r tlw
I 1 I
TP
Fig. 4. Means and standard deviations of Nl-P2 amplitudes. MW: Meaningful words condition; NW: Nonsense words condition; TP: Tone pips condition. Filled rectangles show left NlLP2 amplitudes and empty rectangles show right. Pluses (+) show significant differences between neighbowing values of in&a-hemispheric pairs (MW vs NW, NW vs TP). Asterisks (*) show significant differences between MW and TP of intra-hemispheric pairs. +. *: p < 0.05; z, l: p c 0.01 by dependent t-test. l
the left and right hemispheres (dependent t-test, df: 11, t > 2.20 or t > 3.11) but the comparison between NW and TP revealed that only 2 out of 6 showed significant differences (dependent t-test, df : 11, t > 2.20 or t > 3.11). Comparing the left peak amplitudes with the right, no significant difference in the three conditions was found. The R-values of MW, NW and TP, calculated on the basis of NlLP2 amplitude, were 0.49 + 0.06, 0.5 k 0.04, 0.49 k 0.03 in the frontal region, 0.52 f 0.07, 0.52 f 0.05, 0.51 + 0.04 in the central region and 0.54 + 0.04, 0.5 + 0.07, 0.52 f 0.06 in the parietal region respectively, showing no significant differences among the three conditions. 3.2. Correlational
analysis
(i) Whole sweep time: The mean and s.d. of Fisher’s 2 transformed values of 12 subjects, calculated on the basis of the right and left evoked potentials of homologous regions, are shown in fig. 5. The two-way analysis of variance (task x subjects) were performed on each region. There were significant dif-
Z-va
lue
3.0.
+
2.0.
e t
8
f ..i
-l-
e
-
t
iB
B
-
El
_1.0.
- -
0. FRONTAL
CENTRAL
PARIETAL
Fig. 5. Means and standard deviations of correlation coefficients (Z transformed and right AEPs in 12 subjects. 0: meaningful words condition; 5: Nonsense words Tone pips condition. Asterisks show significant differences between MW and TP. significant difference between MW and NW. +, *: p c 0.05; :: p c 0.01 by Ryan’s
values) of left condition; 8 : A plus shows method.
ferences as a function of task in the frontal region I&(2,33) = 3.66, p < 0.05) and in the central region (F(2,33) = 10.27, p < 0.01). Comparing the three conditions in the frontal and central regions by Ryan’s method, there were significant differences between MW and TP(f = 2.62, df : 33, p -C 0.05) in the frontal region, and between MW and NW(t = 2.83, df : 33, p < 0.05) and between MW and TP( t = 4.33, df : 33, p < 0.01) in the central region. In all cases the significant differences indicated that there was a great dissimilarity between the left and right hemispheric waveforms in the meaningful words condition. (ii) Waveform similarity curve: Fig. 6 shows the waveform similarity curves for the three conditions in the frontal and central regions where significant differences have been found on an examination of the whole sweep time. The curves show the mean value of Fisher’s 2 scores for the 12 subjects in each successive time period. Asterisks above the abscissa indicate the latencies
142
H. Ninom~yo and T. Ikeda / A waueJorm dissimihrrty of AEP hv the memrngful
nouns
Z-value
FRaIlAL 2.90 t.9941 2.50 t.987) 2.10, l.971) 1.70, (.935) 1.30 t.862)
....... ..... ............ ......................... ................
(.79-$ 66
ll0
.........
2;o
TIME 3bo
380
4;o
msu
z -value
(
2.45. .985)
2.00' C.964)
(
1.55' ,914) . =., / -.__ -..* ,.aL"-'-.__.__ _.J -...* __._a----
1.10 ( ,801) 0.65. ( ,572)
............................................. ............................................. .......................................................... .......................................................... TI.ME 6b
li0
22-o
360
se0
SC
Fig. 6. Waveform similarity curves drawn on the averaged Z values of 12 subjects. - - -: Meaningful words condition; -.-‘-; Nonsense words condition; -: Tone pips condition. :: ranges of p -C 0.05; **: Ranges of p < 0.01 by the two-way analysis of variance. Notice that a scale of the *. ordinates (Z value) is different in two regions.
H. Ninomiyu and T. Ikeda
/ A wweform
dissimilarity
of AEP by the meaningful nouns
143
where significant differences were found (by the two-way analysis of variance, task factor; F(2,33) > 3.30 or F(2,33) > 5.34). The significant differences existed at about latencies of 160 and 260 msec in the frontal region and at about 160 and 350 msec in the central region. Around these latencies, the AEPs between the left and right homologous regions in the meaningful words condition were more dissimilar than those in other two conditions. 3.3. Relationship
between Nl -P2 amplitudes
and correlation coefficients
When the correlation coefficients of three conditions between Nl-P2 amplitudes and correlation coefficients of the whole sweep time (C.C.) that were obtained on the basis of the left/right evoked potentials were calculated for each region, there were no significant correlations between NlLP2 amplitudes and C.C. in each condition in each region (range of coefficients (r) is - 0.35 to 0.16). On the other hand, when the correlation coefficients between Nl-P2 amplitudes and C.C.s were calculated intra-individually, 5 of 12 subjects showed significant correlations while 7 of 12 did not. The mean and s.d. of Z-transformed correlation coefficients between Nl-P2 amplitudes and C.C.s for 12 subjects was 0.72 (r = 0.62) f 0.47.
4. Discussion Hemispheric asymmetry in language processing was investigated using auditory evoked potentials as the index for neurophysiological activity. Matched series of tone pips, nonsense words (PIPI, PEPE) and meaningful words in two syllables were used as stimuli. It is well known that the physical properties of a stimulus influence the waveform of evoked potentials, but in general the experimetal control of the physical property is difficult when meaningful words are used as stimuli. In the Japanese language, there are many meaningful nouns in two syllables in which the syllabic stress of the first syllable is almost identical and thus the control of the physical properties can be achieved relatively easily. In the present study, the physical properties of the meaningful words and of the nonsense words were able to be matched. In addition, the experimental procedure was used which ensured that the stimuli were attended and comprehended by the subjects. At the same time the records were not contaminated by motor activities as the subjects were instructed to press a button only after hearing a block of 5 successive stimuli. In previous studies of brain asymmetry using auditory evoked potentials, some authors (Morrell and Salamy, 1971; Matsumiya et al., 1972; Wood and Goff, 1971) have reported the enhancement of peak amplitudes over the left hemisphere whereas other authors (Friedman et al., 1975; Galambos et al., 1975; Grabow et al., 1980) have reported negative results. Such contrary
144
H. Ninomrya and T Ikeda / A waoeform dmimilarity
of A EP by the meuningful nouns
findings could suggest that a simple measurement of amplitudes is not a suitable index for brain asymmetry. Such a hypothesis was confirmed in the present study, where Nl-P2 amplitudes in the three conditions showed no significant asymmetry. The variability of peak amplitudes is large when compared inter-individually and was relatively large even in intra-individual comparison when the compared evoked potentials were recorded in the specified time interval (Roth et al., 1975). It may be possible that any significant differences are concealed by the large inter-individual variability or by the fact that comparisons have been made using evoked potentials averaged over fairly large time periods. In the present study, one of the methods adopted to reduce such variability in evoked potentials was to calculate a coefficient on the basis of evoked potentials recorded simultaneously within the same subject. However, the result from the analysis of a R-value showed no significant asymmetry. To calculate the correlation coefficient between the left and right evoked potentials recorded simultaneously is also one of the methods that reduces the variabilities. In the present study, correlation coefficients between the left and right evoked potentials of the whole sweep time and waveform similarity curves have been used to test asymmetry. Examination of the correlation coefficients for the whole sweep time showed significant differences between the meaningful words condition and the tone pips condition in the frontal region (by Ryan’s method, t = 2.62, df: 33, p < 0.05) and between the meaningful words condition and the nonsense words condition (t = 2.83, df : 33, p < O.OS), and the tone pips condition (t = 4.33, df : 33, p < 0.01) in the central region. Further examination of the frontal and central regions using a waveform similarity curve showed that the correlation coefficients of the right and left evoked potentials in the meaningful words condition was significantly lower than those in other conditions at two latencies, namely at about 160 msec in both the frontal and central regions, and later at about 260 msec in the frontal region and at about 350 msec in the central region (by ANOVA, p < 0.05 or p -c 0.01). The stimuli used in the present study were nonsense and meaningful words in two syllables and tone pips whose physical properties were matched to be as similar as possible. While the tone pips was a mechanical sound, both the nonsense and meaningful words were spoken by the human voice so that both had similar physical properties. However, while the meaningful words which has semantic properties required attention to whole words, nonsense words had no semantic property and it was not necessary for the subjects to attend to each successive syllable. Differences in peak amplitudes and waveforms among three conditions can be attributed to such factors. Although the physical properties of the stimuli in three conditions, particularly in the nonsense words and meaningful words conditions, were matched, the smaller peak amplitudes were recorded in the meaningful words condition,
H. Ninomiycr and T. Ikeda / A waueformdissimilarity
of AEP by the meaningful
mum
145
especially when Nl was used for the measurement (fig. 3). Such a finding could be due to a number of reasons including: (1) A more gradual stimulus onset causes less precise time-locking; (2) the variability among the meaningful words is relatively large; (3) the process of recognizing and classifying the meaningful words is complex, and it causes less synchrony than that of simple stimuli; (4) one of the authors investigated the relation between the subject’s attention to the stimulus and Nl(Ikeda, 1973) and reported that the early negative component (Nl) became smaller when the stimulus was delivered at the point where EEG changed in negativity (eg. CNV). In the present study, the difficulty of the task in the meaningful words condition is much greater than that in other conditions, and it may be suggested that because the subjects were required to attend intensively until all 5 successive stimuli had been delivered, a negative shift may have occurred. Such an interpretation seems possible as one of the reasons. It has been reported that there was a relationship between a peak amplitude and a correlation coefficient (Callaway, 1975). In the present study, the Nl-P2 amplitudes of the meaningful words condition were smaller than those of other conditions, and correlation coefficients of the meaningful words condition were lower than those of others. But, if the correlation coefficients between Nl-P2 amplitudes and C.C.s were calculated intra-sessionally, there was no significant correlation between these two values. When the relationship between Nl-P2 amplitudes and C.C.s were examined intra-individually, 5 of 12 subjects showed a significant correlation between NlLP2 amplitudes and C.C.s, but the other 7 subjects did not. A possibility can not be fully omitted that the low correlation coefficients of the meaningful words condition was due to the small amplitudes, however, the main reason for the low correlation coefficients between the left and right evoked potential in the meaningful words condition would be the same as the reason that caused the small Nl-P2 amplitudes, i.e. the reasons (3) and/or (4) mentioned above that were essential to process the meaningful nouns. The grand averages of evoked potentials for three conditions were compared with the correlation coefficient data obtained from the analysis of the waveform similarity curve in order to describe in detail the way in which evoked potentials changed as a function of changes in the nature of the stimuli. The first dissimilarity noted in the segmental waveform analysis (at about 160 msec) corresponded to P2, and the negative shift following P2 of the grand averaged AEPs in the meaningful words condition. The second dissimilarity (at about 260 msec in the frontal region and at about 350 msec in the central region), taking the latencies into account, might correspond to P300 that was suggested to be concerned with the resolution of uncertainty (Sutton, 1979) and to P2 and the negative shift following P2 which might be induced by the second syllable (the mean onset of second syllable was 113 msec from the onset of the first syllable) in the meaningful words condition. It was suggested that a
simple measurement of conventional AEP components might be an inadequate method to investigate brain asymmetry. But, in any case, it is interesting to note that the two sections where more dissimilarities between the left and right evoked potentials were shown for the waveform similarity curve analysis in the meaningful words conditions corresponded to the segments in which the grand averages of the auditory evoked potential in the meaningful words condition shifted to negativity. The two sections where significant differences existed for examination of waveform similarity curve analysis appeared before the response to the meaningful words used in the present study, i.e. the mean time of 8 subjects to classify the meaningful words was 453 msec from the onset of the second syllable (unpublished data). The time required from the initiation of the motor potential to the response pulse was 80-100 msec (Bitter, Simson and Vaughan, Jr., 1972). Thus, these two sections existed before the response, even when the part of the record included the time required for a response selection. The relationship between these two sections and the physcholigal process of recognition and classification of the meaningful words is not entirely clear. Howewer, the analysis showed that the two sections, the early part and later part, where the correlation coefficients of the meaningful words condition were significantly lower than those of others, may reflect the neurophysiological asymmetry of each hemisphere in language processing.
References Broca, P. (1861). Remarques sur le siege de la faculte du langage articule; suivie d’une observation d’aphasie. In: Ohashi, H. Clinical Neuropsychology. Igaku Shoin: Tokyo, 22. Brown, W., Marsh, J. and Smith, J. (1973). Contextual meaning effects on speech evoked potentials. Behavioral Biology, 9, 755-761. Brown, W., Marsh, J. and Smith, J. (1976). Evoked potential waveform differences produced by the perception of different meanings of ambiguous phrase. Electroencephalography and Clinical Nemophysiology, 41, 113-123. Buchsbaum, M. and Fedio, P. (1970). Hemispheric differences in evoked potentials to verbal and nonverbal stimuli in the left and right visual fields. Physiology and Behavior, 5, 207-210. Callaway, E. (1975). Brain Electrical Potentials and Individual Psychological Differences. Grune & Stratton: New York, 153-169. Donchin, E. (1979). Event-related brain potentials: A tool in the study of human information processing. In: Begleiter, H. (Ed.). Evoked Brain Potentials and Behavior. Plenum Press: New York, 13-88. Friedman, D., Simson, R., Ritter, W. and Rapin, I. (1975). Cortical evoked potentials elicited by real speech words and human sounds. Electroencephalography and Clinical Neurophysiology, 38, 13-19. Galambos, R., Benson, P.. Smith, T., Schulman-Galambos, C. and Osier, H. (1975). On hemispheric differences in evoked potentials to speech stimuli. Electroencephalography and Clinical Neurophysiology, 39, 279-283. Geschwind, N. (1970). The organization of language and the brain. Science, 170. 940-945.
H. Ninomi)?a and T. Ikeda / A waue/orm dissrmilarity
of AEP by the meaningful nouns
147
Geschwind, N. and Levitsky, W. (1968). Human brain: Left-right asymmetries in temporal speech region. Science, 161, 186-187. Grabow, J.D.. Aronson, A.E., Rose, D.E. and Greene, K.L. (1980). Summated potentials evoked by speech sounds for determining cerebral dominance for language. Electroencephalography and Clinical Neurophysiology, 49, 38-47. Harmony, T., Richard, J.. Ostero. G.. Fernandez, G.. Llorente. S. and Valdes, P. (1973). Symmetry of the visual evoked potential in normal subjects. Electroencephalography and Clinical Neurophysiology, 35, 237-240. Ikeda, T. (1973). Reaction times and early negative components of evoked potentials. Folia Psychiatrica et Neurologica Japonica. 27, 241-250. Kimura, D. (1967). Functional asymmetry of the brain in dichotic listening. Cortex, 3, 163-178. Matsumiya, Y., Tagliasco, V., Lombroso, C.T. and Goodglass, H. (1972). Auditory evoked response: Meaningfulness of stimuli and interhemispheric asymmetry. Science, 175, 790-792. Molfese, D., Papanicolaou, A.. Hess, T. and Molfese, V. (1979). Neuroelectrical correlates of semantic process. In: Begleiter, H. (Ed.). Evoked Brain Potentials and Behavior. Plenum Press: New York, 89-106. Morrell, L. and Salamy, J. (1971). Hemispheric asymmetry of electrical responses to speech stimuli. Science, 174, 164-166. Nakahara. D. and Ikeda, T. (1981). A method for estimation of waveform similarity in averaged evoked potentials (in Japanese). Journal of the Physiological Society of Japan, 43, 1 l-13. Ritter, W., Simson, R. and Vaughan, H.G. Jr. (1972). Association cortex potentials and reaction time in auditory discrimination task. Electroencephalography and Clinical Neurophysiology. 33, 547-555. Roth, W.T., Kopell, B.S., Tinklenberg, J.R., Huntsberger, G.E. and Kraemer, H.C. (1975). Reliability of the contingent negative variation and the auditory evoked potential. Electroencephalography and Clinical Neurophysiology, 38, 45-50. Sutton, S. (1979). P300 - Thirteen years later. In: Begleiter, H. (Ed.). Evoked Brain Potentials and Behavior. Prenum Press: New York, 107-126. Wada, J.A. and Rasmussen, T. (1960). Intracarotid injection of sodium amytal for the lateralization of cerebral speech dominance: Experimental and clinical observation. Journal of Neurosurgery, 17, 266-282. Wernicke, C. (1874). Der Aphasische Symptomencomplex. In: Ohashi, H. Clinical Neuropsychology. Igaku Shoin: Tokyo, 47-53. Wood, C.C. and Goff, W.R. (1971). Auditory evoked potentials during speech perception. Science, 173, 1248-1251.
Biological Psychology 18 (1984) 133-147 North-Holland
A WAVEFORM POTENTIALS NOUNS Hideaki
NINOMIYA
Department Accepted
DISSIMILARITY OF AUDITORY EVOKED INDUCED BY THE STIMULATION OF MEANINGFUL
of Psychiatry, for publication
* and Terutika
IKEDA
**
Miyaraki
Medical College, Mlyaraki,
31 August
1983
Japan
Hemispheric asymmetry in language processing was investigated using an auditory evoked potential as the index for neurophysiological activity. Subjects were given the task of recognizing and classifying typical Japanese nouns, responding to nonsense syllables and responding to tone pips. Fourteen subjects participated and the AEPs of 12 subjects from the frontal, central and parietal regions were adopted for subsequent analyses. Correlation coefficients between the right/left AEPs from homologous scalp regions and a waveform similarity curve analysis that can indicate the similarity of two AEPs segmentally and successively, were adopted to provide the intra-individual symmetry index. Results were as follows: (1) The waveforms between the right/left AEPs induced by meaningful nouns were significantly dissimilar than those induced by nonsense syllables and tone pips; (2) the dissimilarities occurred mainly at 160 msec and 350 msec from the onset of the first syllable on the frontal and central regions. These findings may reflect the hemispheric asymmetry of neurophysiological activity in language processing.
1. Introduction The assumption that each hemisphere of the brain has different functions, and that the left hemisphere plays the dominant role in language processing was suggested by early clinical, anatomical and psychophysiological findings (e.g. Broca, 1861; Wernicke, 1874; Wada and Rasmussen, 1960; Kimura, 1967; Geschwind and Levitsky, 1968; Geschwind, 1970). In the neurophysiological field, to test the asymmetrical functions of the left/right hemispheres in language processing, many recent studies have been using the evoked potentials as an index. * Address Medical ** We wish Maxwell
requests for reprints to: Hide&i Ninomiya, Department of Psychiatry, Miyazaki College, 5200 Kihara, Kiyotake-cho, Miyazaki-gun, Miyazaki, Japan. to express our thanks to Ms. Y. Ibusuki and J. Ono for their assistance, to Mrs. G. and Mr. D. Nakahara for their helpful discussion.
0301-0511/84/$3.00
0 1984, Elsevier Science Publishers
B.V. (North-Holland)
134
H. NI~OWIIJYIand T. Ikeda / A waveform dissimilarity
of AEP b.v the me~n~nRful noun.s
Buchsbaum and Fedio (1970) first reported that the waveforms of visual evoked potentials that were induced by verbal and non-verbal stimuli were more dissimilar in the left hemisphere, as compared with the right, suggesting that hemispheric differences in the averaged evoked potentials were consistent with the hypothesis that each hemisphere of the brain had different roles. Investigations using auditory evoked poentials (AEPs) reported the same in the early 1970’s. Morrell and Salamy (1971) and Matsumiya, Tagliasco. Lombroso and Goodglass (1972) reported that the peak amplitude of Nl(Morrel1 and Salamy, 1971) or ‘ W’ wave (Matsumiya et al., 1972) in the left temporo-parietal area was larger than those in the right. Wood and Goff (1971) divided the AEP into two parts, i.e. the preresponse interval and the response interval, reporting that the amplitude of each sampling point of a task requiring the analysis of linguistic information was significantly different in the preresponse interval from those of a task which did not require the analysis of linguistic information. On the other hand, some studies failed to confirm the hypothesis of the hemispheric asymmetry (Friedman, Simson, Ritter and Rapin. 1975; Galambos, Benson, Smith, Schulman-Galambos and Osier, 1975; Grabow, Aronson, Rose and Green, 1980). Friedman et al. (1975) closely examined the experimental design and the statistical analysis of early studies, and concluded that the data demonstrating the hemispheric asymmetry of evoked potentials were far from convincing. Since Grabow et al. (1980) who replicated the study of Morrell and Salamy (1971) were unable to find the hemispheric differences, it could be suggested that the use of a simple measurement of the peak amplitudes was inadequate as the test for asymmetry. Brown, Marsh and Smith (1973. 1976) also used a correlational method similar to that of Buchsbaum and Fedio (1970). They compared the AEPs induced by various meaningful words and found that the words which were similar in sound but not in meaning, produced more dissimilar waveforms in the left hemisphere than in the right. They also suggested in their 1976 paper that their method of using the coefficient of correlation calculated over the entire evoked potential did not provide any indication of response epochs or components. Donchin (1979) suggested that since many cognitive variables might affect evoked potentials, it might be more appropriate to use multivariate methods of analysis. Molfese, Papanicolaou, Hess and Molfese (1979), using a principal component analysis, reported that on three of the factors, there was a significant effect as a function of meaning and that in each case the effect was noted over a short period. In most studies mentioned above, the compared evoked potentials were recorded successively at different times. Roth, Kopell, Tinklenberg, Huntsberger and Kraemer (1975) reported that investigating the test/re-test reliability of the averaged evoked potentials of the same person at successive times, the median correlation coefficient was formed to be 0.59. In other words, there
H. Ninomtya
and T. Ikeda / A waveform dissimilarrty of AEP by the meaningful
mum
135
was a moderate amount of intra-individual consistency. It is well known that the inter-individual variability of evoked potentials is large. One of the methods of eliminating the variability in evoked potentials caused by changes over some time, or differences between individuals, is to calculate a coefficient based on the data from the same individual on the basis of evoked potentials recorded simultaneously. Harmony, Ricard, Ostero, Fernandez, Llorente and Valdes (1973) reported that the evoked potentials of bilateral homologous ‘regions which were recorded simultaneously showed quite similar waveforms. For this reason coefficients calculated in this way would provide the index for symmetry. Perhaps the meaningful words would be sensed in both hemispheres but the recognition or classification of them would be processed mainly in the left hemisphere. Thus, it can be hypothesized that the synchronization of the left and right evoked potentials elicited by the meaningful words would be less than that elicited by pure tones. In the present study, the subjects were given the tasks of recognizing and classifying some typical Japanese nouns in two syllables, as well as responding to nonsense syllables and to tone pips, and a correlational analysis was adopted in two ways: (1) Correlation coefficients have been calculated on the basis of evoked potentials, recorded simultaneously, of the left and right homologous regions to provide the intra-individual symmetry index; (2) as Molfese et al. (1979) reported that the effect of meaning occurred at short periods, there was a possibility that the left and right desynchronization presented only for a short time during the processing of a meaningful stimulus. If such was the case, the analysis of the whole sweep time was unlikely to detect asymmetries. Therefore, the data were also subjected to an analysis in segments by a waveform similarity curve method that is a modified correlational analysis (Nakahara and Ikeda, 1981).
2. Methods 2.1. Subjects and setting Fourteen paid subjects (two females and twelve males, age 23-26) participated in the experiment. All were right handed and audiologically normal. They were seated on a reclining chair, located 1 m in front of a loud speaker in a sound-damped and electrostatically shielded room which was dimly lit (40 lux). Ag/AgCl disc electrodes were placed bilaterally on the scalp at the frontal, central and parietal regions according to the ten-twenty system. The linked-ear electrodes were used for a reference. EEGs were amplified by Nihon Koden AB-620G bioelectric amplifier (0.3 set in time constant and 100 Hz high-cut filter), and were recorded on a Sony UFR-70760 data recorder. The arousal
136
H. Ninomiya
and T. lkeda / A waveform dissmilarit~y of AEP by the meaningful nouns
level and eye movements were monitored by an EEG paper recorder. Two out of 14 records were omitted from subsequent analysis because of an excessive contamination by eye movements. 2.2. Stimulus The real voice of a female volunteer and tone pips were used for the stimuli, and two kinds of speech were used; one was meaningful words in two syllables, and the other was nonsense words in two syllables (PIPI, PEPE). An example of each stimulus is shown in fig. 1. Each verbal stimulus lasted for about 200 msec and the intensity of all the stimuli was adjusted to be as similar as possible. The mean onset of the second syllable of the meaningful words was 113 + 43 msec after the onset of the first syllable. As for the tone pips, two kinds of stimuli were used; one was 300 Hz, 65 dB SPL, 200 msec duration, and another was 600 Hz, 65 dB SPL, 200 msec duration. 2.3. Procedure There were three conditions: (a) Meaningful words condition (MW); (b) nonsense words condition (NW); (c) tone pips conditions (TP). Each condition consisted of 20 blocks of 5 successive stimuli. Examples of each block are shown in fig. 2. In the meaningful words condition, the words in two syllables (consonantvowel-consonant-vowel) were used as stimuli. In 20 blocks, there were two types of blocks, one consisted of 5 words from the same class, and the other consisted of 4 words from the same class mixed with the word from the
ZE 100 msec
300 Hz TONE PIPS Fig. 1. An example of stimuli in three conditions. Each stimulus is put into Sanei Signal Processor and drawn out by the Watanabe (window).
X-Y
recorder. The meaningful
word shown in the fig. is kaze
H. Ninomiya
(,)
and T. Ike-da / A waue/orm dissimilarity
k mo
(A\
nami
ki
of AEP
by the meaningful nouns
137
button press
I
nr (B)
(III)
(A)
pipi
I
pipi
pi i
300
300
300
i
epe
300
r
r
‘7sec
f
Fig. 2. An example of the blocks where a button press is required and where not required. (I) meaningful words condition, (II) nonsense words condition, (III) tone pips condition. (I) (A) all nouns are the class of natural phenomena, (B) mixed with plants (taka is bamboo). Meaningful nouns shown in English are kaze (window), kiri (fog), kumo (cloud), yuki (snow) and nami (wave). t-- : Beginning of the next block.
different class. The classes used were as follows: The natural phenomena; parts of body; personal ornaments; animals; fishes; birds; and plants. To check that the stimuli had been attented to and comprehended, the subjects were instructed to push a button at the end of each block if they thought the block had consisted of 5 words from the same class. In the nonsense words condition, there were the block of 5 PIPIs and that of 4 PIPIs with one PEPE and in the tone pips condition, the block of five 300 Hz tone pips and that of four 300 Hz with one 600 Hz. They were also instructed to push the button if 5 successive stimuli (i.e. one block) were identical. As the finger lifting or button press occurred only after a block of stimuli had been presented, the motor potential that contaminated the late component of auditory evoked potentials could be ignored. The nonsense words condition was used because a possibility existed that a human voice even without meaning might induce an asymmetry in the evoked potentials (Morrell and Salamy, 1971). Six subjects participated in the experimental sequence of (a)-(b)-(c) and other 6 subjects in (c)-(b)-(a) to counterbalance for order effect. 2.4. Data analysis The EEG of each subject was averaged over 100 times by Sanei Signal Processor 7T07A from a trigger pulse to 510 msec with a sampling rate of 1 point per 2 msec. The averaged potentials were written out as an analogue record by the Watanabe X-Y recorder, and also were punched out in a digital form by Kasio model 501 typuter. Punch tapes were processed by Canon BX-1 microcomputer for further analysis.
(a) Component analysis The NlLP2 peak-to-peak amplitude were measured, and R-values were calculated by the same method as that of Matsumiya et al. (1972). A R-value greater than 0.5 indicated that the left evoked potential was greater than the right, and smaller than 0.5, the right greater than the left. In the present paper, the Nl-P2 peak-to-peak amplitude was used to calculate the R-value. (b) Correlational analysis (i) Whole sweep time (510 msec): Pearson product moment correlation coefficients of evoked potentials from the left and right homologous scalp regions were calculated by Canon BX-1 on the basis of digitised punch-tapes. The calculated values indicated the degree of resemblance between the waveforms of the left and right evoked potentials. (ii) Waveform similarity curve: Essentially this method was a correlational technique. There were two methodological advantages, one was that the optimal sampling length (window) to calculate a correlation coefficient could be taken out, and another was that the window moved from the trigger point to the end of sweep time step by step on the sampling time. Therefore, it could indicate the degree of the resemblance between two evoked potentials segmentally and successively. Details of this method have been fully described in a previous paper (Nakahara and Ikeda, 1981). In the present paper, the sampling length of the waveform similarity curve was 118 msec (sampling point = 60), therefore correlation coefficients of two evoked potentials were calculated from O-118 msec to 3922510 msec successively (the total number of correlation coefficients was 197). The calculated value was dotted at the center of each sampling length (59, 61, 63, . . ., 451 msec).
3. Results As the mean percents of correct responses were not different among the three conditions (> 99%) it was suggested that each subject understood the task and attended to the stimuli. The grand averages of auditory evoked potentials of 12 subjects for each region at each condition are shown in fig. 3. Comparing Nl and P2 amplitudes among the three conditions, the Nl amplitudes of nonsense words and tone pips conditions were larger than those of the meaningful words condition and there was not much difference in the P2 amplitudes among the three conditions. 3. I. Component analysis The mean and s.d. of Nl-P2 amplitudes are shown in fig. 4. All the NlLP2 amplitudes of MW were significantly lower than those of NW and TP both in
: :Ij
H. Nlnomiyu
:
‘...,.
..
and T. Ikedcr / A waveform
dissimilarity
o/A EP by the meaningful
nouns
139
140
H. Ninomiya
and T. Ikeda / A waueform dissimilarity
FRONTAL
of A EP by the memingful
CENTRAL
nouns
PARIETAL
4v
10
10
10
: ::
:* :;
5
0
II
.
II
r-m
NW
TP
:
I
:+ :I 5
5
0
_
i rlw
TP
.
.r tlw
I 1 I
TP
Fig. 4. Means and standard deviations of Nl-P2 amplitudes. MW: Meaningful words condition; NW: Nonsense words condition; TP: Tone pips condition. Filled rectangles show left NlLP2 amplitudes and empty rectangles show right. Pluses (+) show significant differences between neighbowing values of in&a-hemispheric pairs (MW vs NW, NW vs TP). Asterisks (*) show significant differences between MW and TP of intra-hemispheric pairs. +. *: p < 0.05; z, l: p c 0.01 by dependent t-test. l
the left and right hemispheres (dependent t-test, df: 11, t > 2.20 or t > 3.11) but the comparison between NW and TP revealed that only 2 out of 6 showed significant differences (dependent t-test, df : 11, t > 2.20 or t > 3.11). Comparing the left peak amplitudes with the right, no significant difference in the three conditions was found. The R-values of MW, NW and TP, calculated on the basis of NlLP2 amplitude, were 0.49 + 0.06, 0.5 k 0.04, 0.49 k 0.03 in the frontal region, 0.52 f 0.07, 0.52 f 0.05, 0.51 + 0.04 in the central region and 0.54 + 0.04, 0.5 + 0.07, 0.52 f 0.06 in the parietal region respectively, showing no significant differences among the three conditions. 3.2. Correlational
analysis
(i) Whole sweep time: The mean and s.d. of Fisher’s 2 transformed values of 12 subjects, calculated on the basis of the right and left evoked potentials of homologous regions, are shown in fig. 5. The two-way analysis of variance (task x subjects) were performed on each region. There were significant dif-
Z-va
lue
3.0.
+
2.0.
e t
8
f ..i
-l-
e
-
t
iB
B
-
El
_1.0.
- -
0. FRONTAL
CENTRAL
PARIETAL
Fig. 5. Means and standard deviations of correlation coefficients (Z transformed and right AEPs in 12 subjects. 0: meaningful words condition; 5: Nonsense words Tone pips condition. Asterisks show significant differences between MW and TP. significant difference between MW and NW. +, *: p c 0.05; :: p c 0.01 by Ryan’s
values) of left condition; 8 : A plus shows method.
ferences as a function of task in the frontal region I&(2,33) = 3.66, p < 0.05) and in the central region (F(2,33) = 10.27, p < 0.01). Comparing the three conditions in the frontal and central regions by Ryan’s method, there were significant differences between MW and TP(f = 2.62, df : 33, p -C 0.05) in the frontal region, and between MW and NW(t = 2.83, df : 33, p < 0.05) and between MW and TP( t = 4.33, df : 33, p < 0.01) in the central region. In all cases the significant differences indicated that there was a great dissimilarity between the left and right hemispheric waveforms in the meaningful words condition. (ii) Waveform similarity curve: Fig. 6 shows the waveform similarity curves for the three conditions in the frontal and central regions where significant differences have been found on an examination of the whole sweep time. The curves show the mean value of Fisher’s 2 scores for the 12 subjects in each successive time period. Asterisks above the abscissa indicate the latencies
142
H. Ninom~yo and T. Ikeda / A waueJorm dissimihrrty of AEP hv the memrngful
nouns
Z-value
FRaIlAL 2.90 t.9941 2.50 t.987) 2.10, l.971) 1.70, (.935) 1.30 t.862)
....... ..... ............ ......................... ................
(.79-$ 66
ll0
.........
2;o
TIME 3bo
380
4;o
msu
z -value
(
2.45. .985)
2.00' C.964)
(
1.55' ,914) . =., / -.__ -..* ,.aL"-'-.__.__ _.J -...* __._a----
1.10 ( ,801) 0.65. ( ,572)
............................................. ............................................. .......................................................... .......................................................... TI.ME 6b
li0
22-o
360
se0
SC
Fig. 6. Waveform similarity curves drawn on the averaged Z values of 12 subjects. - - -: Meaningful words condition; -.-‘-; Nonsense words condition; -: Tone pips condition. :: ranges of p -C 0.05; **: Ranges of p < 0.01 by the two-way analysis of variance. Notice that a scale of the *. ordinates (Z value) is different in two regions.
H. Ninomiyu and T. Ikeda
/ A wweform
dissimilarity
of AEP by the meaningful nouns
143
where significant differences were found (by the two-way analysis of variance, task factor; F(2,33) > 3.30 or F(2,33) > 5.34). The significant differences existed at about latencies of 160 and 260 msec in the frontal region and at about 160 and 350 msec in the central region. Around these latencies, the AEPs between the left and right homologous regions in the meaningful words condition were more dissimilar than those in other two conditions. 3.3. Relationship
between Nl -P2 amplitudes
and correlation coefficients
When the correlation coefficients of three conditions between Nl-P2 amplitudes and correlation coefficients of the whole sweep time (C.C.) that were obtained on the basis of the left/right evoked potentials were calculated for each region, there were no significant correlations between NlLP2 amplitudes and C.C. in each condition in each region (range of coefficients (r) is - 0.35 to 0.16). On the other hand, when the correlation coefficients between Nl-P2 amplitudes and C.C.s were calculated intra-individually, 5 of 12 subjects showed significant correlations while 7 of 12 did not. The mean and s.d. of Z-transformed correlation coefficients between Nl-P2 amplitudes and C.C.s for 12 subjects was 0.72 (r = 0.62) f 0.47.
4. Discussion Hemispheric asymmetry in language processing was investigated using auditory evoked potentials as the index for neurophysiological activity. Matched series of tone pips, nonsense words (PIPI, PEPE) and meaningful words in two syllables were used as stimuli. It is well known that the physical properties of a stimulus influence the waveform of evoked potentials, but in general the experimetal control of the physical property is difficult when meaningful words are used as stimuli. In the Japanese language, there are many meaningful nouns in two syllables in which the syllabic stress of the first syllable is almost identical and thus the control of the physical properties can be achieved relatively easily. In the present study, the physical properties of the meaningful words and of the nonsense words were able to be matched. In addition, the experimental procedure was used which ensured that the stimuli were attended and comprehended by the subjects. At the same time the records were not contaminated by motor activities as the subjects were instructed to press a button only after hearing a block of 5 successive stimuli. In previous studies of brain asymmetry using auditory evoked potentials, some authors (Morrell and Salamy, 1971; Matsumiya et al., 1972; Wood and Goff, 1971) have reported the enhancement of peak amplitudes over the left hemisphere whereas other authors (Friedman et al., 1975; Galambos et al., 1975; Grabow et al., 1980) have reported negative results. Such contrary
144
H. Ninomrya and T Ikeda / A waoeform dmimilarity
of A EP by the meuningful nouns
findings could suggest that a simple measurement of amplitudes is not a suitable index for brain asymmetry. Such a hypothesis was confirmed in the present study, where Nl-P2 amplitudes in the three conditions showed no significant asymmetry. The variability of peak amplitudes is large when compared inter-individually and was relatively large even in intra-individual comparison when the compared evoked potentials were recorded in the specified time interval (Roth et al., 1975). It may be possible that any significant differences are concealed by the large inter-individual variability or by the fact that comparisons have been made using evoked potentials averaged over fairly large time periods. In the present study, one of the methods adopted to reduce such variability in evoked potentials was to calculate a coefficient on the basis of evoked potentials recorded simultaneously within the same subject. However, the result from the analysis of a R-value showed no significant asymmetry. To calculate the correlation coefficient between the left and right evoked potentials recorded simultaneously is also one of the methods that reduces the variabilities. In the present study, correlation coefficients between the left and right evoked potentials of the whole sweep time and waveform similarity curves have been used to test asymmetry. Examination of the correlation coefficients for the whole sweep time showed significant differences between the meaningful words condition and the tone pips condition in the frontal region (by Ryan’s method, t = 2.62, df: 33, p < 0.05) and between the meaningful words condition and the nonsense words condition (t = 2.83, df : 33, p < O.OS), and the tone pips condition (t = 4.33, df : 33, p < 0.01) in the central region. Further examination of the frontal and central regions using a waveform similarity curve showed that the correlation coefficients of the right and left evoked potentials in the meaningful words condition was significantly lower than those in other conditions at two latencies, namely at about 160 msec in both the frontal and central regions, and later at about 260 msec in the frontal region and at about 350 msec in the central region (by ANOVA, p < 0.05 or p -c 0.01). The stimuli used in the present study were nonsense and meaningful words in two syllables and tone pips whose physical properties were matched to be as similar as possible. While the tone pips was a mechanical sound, both the nonsense and meaningful words were spoken by the human voice so that both had similar physical properties. However, while the meaningful words which has semantic properties required attention to whole words, nonsense words had no semantic property and it was not necessary for the subjects to attend to each successive syllable. Differences in peak amplitudes and waveforms among three conditions can be attributed to such factors. Although the physical properties of the stimuli in three conditions, particularly in the nonsense words and meaningful words conditions, were matched, the smaller peak amplitudes were recorded in the meaningful words condition,
H. Ninomiycr and T. Ikeda / A waueformdissimilarity
of AEP by the meaningful
mum
145
especially when Nl was used for the measurement (fig. 3). Such a finding could be due to a number of reasons including: (1) A more gradual stimulus onset causes less precise time-locking; (2) the variability among the meaningful words is relatively large; (3) the process of recognizing and classifying the meaningful words is complex, and it causes less synchrony than that of simple stimuli; (4) one of the authors investigated the relation between the subject’s attention to the stimulus and Nl(Ikeda, 1973) and reported that the early negative component (Nl) became smaller when the stimulus was delivered at the point where EEG changed in negativity (eg. CNV). In the present study, the difficulty of the task in the meaningful words condition is much greater than that in other conditions, and it may be suggested that because the subjects were required to attend intensively until all 5 successive stimuli had been delivered, a negative shift may have occurred. Such an interpretation seems possible as one of the reasons. It has been reported that there was a relationship between a peak amplitude and a correlation coefficient (Callaway, 1975). In the present study, the Nl-P2 amplitudes of the meaningful words condition were smaller than those of other conditions, and correlation coefficients of the meaningful words condition were lower than those of others. But, if the correlation coefficients between Nl-P2 amplitudes and C.C.s were calculated intra-sessionally, there was no significant correlation between these two values. When the relationship between Nl-P2 amplitudes and C.C.s were examined intra-individually, 5 of 12 subjects showed a significant correlation between NlLP2 amplitudes and C.C.s, but the other 7 subjects did not. A possibility can not be fully omitted that the low correlation coefficients of the meaningful words condition was due to the small amplitudes, however, the main reason for the low correlation coefficients between the left and right evoked potential in the meaningful words condition would be the same as the reason that caused the small Nl-P2 amplitudes, i.e. the reasons (3) and/or (4) mentioned above that were essential to process the meaningful nouns. The grand averages of evoked potentials for three conditions were compared with the correlation coefficient data obtained from the analysis of the waveform similarity curve in order to describe in detail the way in which evoked potentials changed as a function of changes in the nature of the stimuli. The first dissimilarity noted in the segmental waveform analysis (at about 160 msec) corresponded to P2, and the negative shift following P2 of the grand averaged AEPs in the meaningful words condition. The second dissimilarity (at about 260 msec in the frontal region and at about 350 msec in the central region), taking the latencies into account, might correspond to P300 that was suggested to be concerned with the resolution of uncertainty (Sutton, 1979) and to P2 and the negative shift following P2 which might be induced by the second syllable (the mean onset of second syllable was 113 msec from the onset of the first syllable) in the meaningful words condition. It was suggested that a
simple measurement of conventional AEP components might be an inadequate method to investigate brain asymmetry. But, in any case, it is interesting to note that the two sections where more dissimilarities between the left and right evoked potentials were shown for the waveform similarity curve analysis in the meaningful words conditions corresponded to the segments in which the grand averages of the auditory evoked potential in the meaningful words condition shifted to negativity. The two sections where significant differences existed for examination of waveform similarity curve analysis appeared before the response to the meaningful words used in the present study, i.e. the mean time of 8 subjects to classify the meaningful words was 453 msec from the onset of the second syllable (unpublished data). The time required from the initiation of the motor potential to the response pulse was 80-100 msec (Bitter, Simson and Vaughan, Jr., 1972). Thus, these two sections existed before the response, even when the part of the record included the time required for a response selection. The relationship between these two sections and the physcholigal process of recognition and classification of the meaningful words is not entirely clear. Howewer, the analysis showed that the two sections, the early part and later part, where the correlation coefficients of the meaningful words condition were significantly lower than those of others, may reflect the neurophysiological asymmetry of each hemisphere in language processing.
References Broca, P. (1861). Remarques sur le siege de la faculte du langage articule; suivie d’une observation d’aphasie. In: Ohashi, H. Clinical Neuropsychology. Igaku Shoin: Tokyo, 22. Brown, W., Marsh, J. and Smith, J. (1973). Contextual meaning effects on speech evoked potentials. Behavioral Biology, 9, 755-761. Brown, W., Marsh, J. and Smith, J. (1976). Evoked potential waveform differences produced by the perception of different meanings of ambiguous phrase. Electroencephalography and Clinical Nemophysiology, 41, 113-123. Buchsbaum, M. and Fedio, P. (1970). Hemispheric differences in evoked potentials to verbal and nonverbal stimuli in the left and right visual fields. Physiology and Behavior, 5, 207-210. Callaway, E. (1975). Brain Electrical Potentials and Individual Psychological Differences. Grune & Stratton: New York, 153-169. Donchin, E. (1979). Event-related brain potentials: A tool in the study of human information processing. In: Begleiter, H. (Ed.). Evoked Brain Potentials and Behavior. Plenum Press: New York, 13-88. Friedman, D., Simson, R., Ritter, W. and Rapin, I. (1975). Cortical evoked potentials elicited by real speech words and human sounds. Electroencephalography and Clinical Neurophysiology, 38, 13-19. Galambos, R., Benson, P.. Smith, T., Schulman-Galambos, C. and Osier, H. (1975). On hemispheric differences in evoked potentials to speech stimuli. Electroencephalography and Clinical Neurophysiology, 39, 279-283. Geschwind, N. (1970). The organization of language and the brain. Science, 170. 940-945.
H. Ninomi)?a and T. Ikeda / A waue/orm dissrmilarity
of AEP by the meaningful nouns
147
Geschwind, N. and Levitsky, W. (1968). Human brain: Left-right asymmetries in temporal speech region. Science, 161, 186-187. Grabow, J.D.. Aronson, A.E., Rose, D.E. and Greene, K.L. (1980). Summated potentials evoked by speech sounds for determining cerebral dominance for language. Electroencephalography and Clinical Neurophysiology, 49, 38-47. Harmony, T., Richard, J.. Ostero. G.. Fernandez, G.. Llorente. S. and Valdes, P. (1973). Symmetry of the visual evoked potential in normal subjects. Electroencephalography and Clinical Neurophysiology, 35, 237-240. Ikeda, T. (1973). Reaction times and early negative components of evoked potentials. Folia Psychiatrica et Neurologica Japonica. 27, 241-250. Kimura, D. (1967). Functional asymmetry of the brain in dichotic listening. Cortex, 3, 163-178. Matsumiya, Y., Tagliasco, V., Lombroso, C.T. and Goodglass, H. (1972). Auditory evoked response: Meaningfulness of stimuli and interhemispheric asymmetry. Science, 175, 790-792. Molfese, D., Papanicolaou, A.. Hess, T. and Molfese, V. (1979). Neuroelectrical correlates of semantic process. In: Begleiter, H. (Ed.). Evoked Brain Potentials and Behavior. Plenum Press: New York, 89-106. Morrell, L. and Salamy, J. (1971). Hemispheric asymmetry of electrical responses to speech stimuli. Science, 174, 164-166. Nakahara. D. and Ikeda, T. (1981). A method for estimation of waveform similarity in averaged evoked potentials (in Japanese). Journal of the Physiological Society of Japan, 43, 1 l-13. Ritter, W., Simson, R. and Vaughan, H.G. Jr. (1972). Association cortex potentials and reaction time in auditory discrimination task. Electroencephalography and Clinical Neurophysiology. 33, 547-555. Roth, W.T., Kopell, B.S., Tinklenberg, J.R., Huntsberger, G.E. and Kraemer, H.C. (1975). Reliability of the contingent negative variation and the auditory evoked potential. Electroencephalography and Clinical Neurophysiology, 38, 45-50. Sutton, S. (1979). P300 - Thirteen years later. In: Begleiter, H. (Ed.). Evoked Brain Potentials and Behavior. Prenum Press: New York, 107-126. Wada, J.A. and Rasmussen, T. (1960). Intracarotid injection of sodium amytal for the lateralization of cerebral speech dominance: Experimental and clinical observation. Journal of Neurosurgery, 17, 266-282. Wernicke, C. (1874). Der Aphasische Symptomencomplex. In: Ohashi, H. Clinical Neuropsychology. Igaku Shoin: Tokyo, 47-53. Wood, C.C. and Goff, W.R. (1971). Auditory evoked potentials during speech perception. Science, 173, 1248-1251.