- Email: [email protected]
Short-term replicability of the mismatch negativity
ELSEVIER
Electroencephalography and clinical Neurophysiology 100 (1996) 549-554
Short-term replicability of the mismatch negativity Caries Escera*,
Caries Grau
Neurod~,namics Laboratory, Department of Psychiatry and Clinical Psychobiology, University c?["Barcehma. P. Vail d'Hebron, 171, 08035 Barcelona. Catalonia. Spain
Accepted for publication: 19 June 1996
Abstract The short-term replicability of the mismatch negativity (MMN) between two recording sessions spaced 2 h apart was evaluated at individual and group levels in a sample of 11 healthy adults. Subjects were presented with a random sequence of 1000 Hz standard (92%) and 1100 Hz deviant (8%) tones while they were reading a book. The N1 and P2 exogenous components to standard tones showed a fairly good individual and group replicability. There were no significant differences in the MMN amplitude and latency between the two sessions in the group of subjects as a whole. The individual replicability of the MMN was not as good as for the NI to standards, reaching significance in only some of the electrodes. This result was, however, similar to that obtained for the NI after deviant tones. The results indicate that the MMN has good replicability at the group level, and further that at the individual level, MMN replicability is similar to that of the N1 to deviants. This suggests that the number of summations should be increased in order to improve the clinical usefulness of the MMN. Keywords: Mismatch negativity (MMN); Sensory memory; Event-related potentials; N1; Auditory function; Test-retest replicability
1. Introduction Infrequent, physically deviant sounds occurring in a series of unattended standard auditory stimuli elicit a component of the event-related potential (ERP) called mismatch negativity (MMN), peaking at 100-250 ms from stimulus onset. The M M N can be obtained by changing any physical attribute of a tone, as well as by changes in more complex sounds (for a review see N~i~it~inen, 1992). Since no M M N can be obtained by presenting the deviant stimulus alone, nor by using long inter-stimulus intervals (ISis), it has been proposed that the physical features of the auditory stimuli are fully analyzed and encoded in neural traces of echoic memory, the M M N being automatically elicited each time the afferent input caused by an auditory stimulus does not match the neuronal representation of the standard stimuli (N~i~it~inen et al., 1989; N~i~it~inen, 1992). Neural generators of M M N have been located in the supratemporal auditory cortex, as revealed by ERPs and magnetic responses, as well as by intracranial recordings * Corresponding author. Tel.: +34 3 4021080, ext. 3047; fax: +34 3 4021584; e-mail: [email protected]
in cats, monkeys and human subjects (reviewed by Alho, 1995). A further contribution to M M N has been located in the right frontal cortex, as revealed by current source density analysis methods (Giard et al., 1990). This frontal activity involved in M M N generation might be associated with an involuntary switching of attention to any discriminable change in the acoustic stimulation (Giard et al., 1990; N~i~it~inen, 1992). The special interest of M M N for clinical applications is that it is elicited independently of the direction of attention, thus making it possible to study auditory discrimination, sensory memory, and involuntary attention in individuals unable or unwilling to cooperate. Among the most promising developments for MMN, the assessment of cognitive brain dysfunction in normal aging (Woods, 1992; Pekkonen et al., 1993), Alzheimer's disease (Pekkonen et al., 1994), and schizophrenia (Javitt et al., 1995), stands out. In addition, M M N may be used to evaluate hearing rehabilitation after cochlear implants (Ponton and Don, 1995), and as a tool in the prognosis of comatose patients (Kane et al., 1993). However, the final clinical utility of M M N will depend on its individual and overall replicability in short-term measures, when diagnostic,
0168-5597/96/$15.00 © 1996 Elsevier Science Ireland Ltd. All rights reserved Pll S0921-884X(96)95633-7
EEP95633
550
C. Escera, C. Grau / Electroencephalography and clinical Neurophysiology 100 (1996) 549-554
prognostic, or follow-up applications are to be taken into account. In the present paper, the short-term replicability and stability of the MMN, at both individual and group levels, will be studied in a group of healthy adults, by comparing the recordings from two sessions spaced 2 h apart. As one of the most promising clinical uses of MMN may be the monitoring of coma state, in order to identify the earliest sign of recovering of consciousness (Kane et al., 1993), the study was designed to fulfill some of its possible requirements. Thus, the recordings were conducted as they may be performed in the intensive care unit, that is, while the subjects were wearing the electrodes in the interviewing period between recording sessions. Further, the mean age of the subjects studied was 35 years, being in the middle of the age-range covering the 89% of the potential population, according to the Traumatic Coma Data Bank report (Vollmer et al., 1991), to whom a possible MMN test for monitoring of coma states might be applied. 2. Methods 2.1. Subjects
Eleven healthy volunteers (4 females), between the ages of 28 and 41 years (mean age 34.5 + 4.0 years) participated in the experiment. None of them had a previous history of neurological or psychiatric problems, nor of alcohol or drug abuse and, unaware of the purpose of the experiment, they gave informed consent to their participation. Subjects were screened in an audiometric test before starting the recording sessions, and all had a threshold below 40 dB SPL. 2.2. Stimuli and procedure
The Stim software and hardware (Neuroscan, Inc., USA) were used to generate pure sine-wave tones of 60 ms, including 5 ms of rise and fall times, which were presented monaurally to the subject's right ear at an intensity of 85 dB SPL through TDH-39 headphones, with a constant ISI (onset-to-onset) of 480 ms. Standard tones were of 1000 Hz in frequency, and were randomly replaced by deviant tones of 1100 Hz with a probability of 0.08. Four different blocks of 500 stimuli were presented in each of the two recording sessions spaced 2 h apart. Each session lasted for about 20 min, the first starting at approximately 1000 h. Subjects sat in a comfortable chair in an electrically shielded and sound-attenuated room, which was dimly lit (340 lux). The subject was instructed to ignore the stimulation and to concentrate on reading a self-selected book. After the first session, the subject was asked to stay in the same room and to sit on the chair, and to try to remain as relaxed as possible. To control possible effects of fluctuations in subject's sleepiness during the
experiment, subjective alertness was evaluated before each recording session by means of a Catalan version of the Stanford Sleepiness Scale (SSS) (Hodes et al., 1973). The EEG (bandpass 0.15-100 Hz) was continuously amplified and digitized at a rate of 250 Hz/channel, from 7 tin scalp electrodes inserted in a nylon cap (ElectroCap, Inc.), according to the 10-20 system: F3, Fz, F4, C3, Cz, C4, and Pz. One supraorbital electrode and another electrode at the canthus of the left eye served to monitor EOG changes. The reference electrode was attached at the tip of nose, and the subject was grounded with an electrode placed between the Fz and Fpz locations. EEG epochs of 390 ms, beginning 40 ms before each stimulus onset, were obtained off-line by a computer, and averaged separately for the standard and deviant tones. Trials exceeding +80 /~V were automatically excluded from the averages, as well as trials containing excessive eye movements, blinks, bursts of muscle activity, amplifier clipping, or other extra-cerebral artifacts. The first 5 trials of each block were also excluded from the averages. After rejection protocols, al least 130 deviant tones in each recording session were averaged for each subject, the mean number of deviant tones averaged being 148 in both sessions. The ERP waveforms were digitally lowpass filtered at 30 Hz. 2.3. Data analysis
The exogenous NI was defined as the most negative peak in the latency window 80-150 ms after stimulus onset, and the P2 as the most positive peak in the 140220 ms window. The MMN was evaluated as the most negative peak in the 130-280 ms latency window in the difference wave obtained by subtracting the ERP to standard tones from that to deviant tones. All amplitude measurements were made against the mean amplitude of the 40 ms pre-stimulus baseline. Two statistical analyses were performed upon the data. First, to test the individual replicability of the MMN, N1, and P2 between the two recording sessions, Pearson's product-moment correlation coefficients (r) between the two sessions were calculated separately for the measurements obtained at three frontal (F3, Fz, F4) and three central (C3, Cz, C4) electrodes. Then, an analysis of variance for repeated measures (electrode x recording session; ANOVA with P values reported after the GreenhouseGeisser correction when appropriate) were applied to study the test-retest replicability at the group level. 3. Results 3.1. Individual replicability 3.1.1. Mismatch negativity Fig. 1 shows the individual difference waves (deviant ERP minus standard ERP) at F4 in the two recording ses-
C. Escera. C. Grau / Electroencephalography and clinical Neurophysiology 100 (1996) 549-554
551
Table 1 Interindividual range and mean peak amplitude ( + SD) in #V and lalency in ms of the MMN at Fz in the two recording sessions Session
1 2
MMN peak (#V)
MMN latency (ms)
Mean _+SD
Largest
Smallest
Mean + SD
Earliest
Latest
-2.6 _+ 1.3 -2.8 _+ 1.4
-5.0 -5.4
-0.6 -0.9
192 + 26.3 195 + 38.0
160 140
228 260
sions for the 11 subjects studied. A large interindividual variation can be seen in the M M N traces. M e a n M M N amplitude and latency, as well as interindividual range, are displayed for the two sessions in Table 1. The individual short-term replicability of the M M N was estimated by the P e a r s o n ' s p r o d u c t - m o m e n t correlation coefficient b e t w e e n the first and second recording sessions, separately for 6 scalp site recordings. Significant correlation coefficients were found only at F4 (r = 0.598, P = 0.04) and C3 (r = 0.657, P = 0.02) for the M M N amplitude. A tendency to a significant short-term replicability o f the M M N ampli-
tude was also found at F3 (r = 0.532, P = 0.075, n.s.) and Fz (r = 0.516, P = 0.086, n.s.). For the M M N latency, the correlation coefficient was significant only at Fz (r = 0.582, P = 0.047). In an additional analysis, after r e m o v i n g the two subjects (d and j) in w h i c h the identification of the M M N peak was less clear (Fig. 1, bottom), the correlation coefficients for M M N amplitude and latency did not reach significance at any electrode. 3.1.2. N I to d e v i a n t t o n e s The same procedure was also used to test the individual
e
I¢t~
,
1
I
+
+
f
J
-5uV Z
k - -
ist. 2nd.
" rV
'
F
V~v '~
v' 5uV
session session
60ms
I
h
i ,~.i
lh
- -
standard
- -
devian[
i
L
Fig. 1. Top: individual difference waves (ERP to deviant tones minus ERP to standard tones) at F4, where MMN amplitude correlation was significant, for the 11 subjects studied (letters from a to k), in the first and second recording sessions. The first session started at 1000 h, and the second session 2 h later. Bottom: examples of individual ERPs to standard and deviant tones (F4 recording, first session) in which subtraction yielded clear (subject h) or less clear (subject d) MMN peak identification (dark area, MMN).
552
C. Escera, C. Grau / Electroencephalography and clinical Neurophysiology 100 (1996) 549-554
ateral (left-hemisphere) dominance (Fig. 3). The two-way analysis of variance did not reveal any significant differences between the recording sessions for either the N1 amplitude (F(1,10)= 0.76, P = 0.403, n.s.) or latency (F(1,10) = 0.97, P = 0.349, n.s.). The P2 deflection had a mean peak amplitude at Cz of 1.2 #V in the first recording session and 1.0 #V in the second, with respective latencies being 168 ms and 165 ms. The P2 was largest at the central electrodes, and at the midline positions (Fig. 3). No significant differences between the recording sessions were found for either the P2 amplitude (F(1,10) = 2.28, P = 0.162, n.s.) or latency (F(1,10) = 0.21, P = 0.660, n.s.).
short-term replicability of the N1 to deviants at Fz and Cz, where the NI amplitude was larger. A significant correlation was found between the N1 amplitude to deviants in the two recording sessions at Fz (r = 0.661, P = 0.019), but not at Cz (r = 0.499, P = 0.098, n.s.). The latency of the N1 to deviants obtained in the two recording sessions did not correlate significantly at any of the electrodes analyzed (r = 0.267, P = 0.412, n.s., at Fz; r = 0.234, P = 0.464, n.s., at Cz). 3.1.3. ERPs to standard tones
Pearson's product-moment correlation coefficient between the first and the second recording sessions at 6 scalp locations were all highly significant for the N1 and P2 amplitudes (ranging from 0.791, P < 0.01, to 0.948, P < 0.001). For the N1 and P2 latencies, the correlation coefficients were significant at all 6 electrodes, except at F3 for N1 latency, and Cz and C4 for P2 latency (ranging from 0.031, n.s., to 0.939, P < 0.001).
3.2.2. Overall replicability o f the M M N
Fig. 2 also shows the grand-mean difference waves obtained by subtracting the ERP to standard tones from those to deviant tones, for the two recording sessions. The scalp distribution of the MMN peak amplitude was different from that of the N1 deflection, with largest amplitudes at frontal electrodes, whereas the N1 was larger at central scalp sites (Fig. 3). The MMN also tended to be larger over the right hemisphere, as opposed to the N1 deflection, which was larger over the left hemisphere (Fig. 3). The MMN peak showed a slight enhancement in its amplitude in the second recording session (Fig. 3). However, a twoway analysis of variance performed with electrode and recording session as factors failed to reveal statistical significance for the factor session (F(1,10) = 0.73, P = 0.412, n.s.), thus showing a high short-term replicability at the group level. The MMN peak latency was also similar in
3.2. Overall replicability 3.2.1. ERPs to standard tones
Grand-average ERPs to standard and deviant tones in the first and second recording sessions are shown in Fig. 2. Standard tones elicited prominent N1 and P2 deflections in both recording sessions. N1 mean peak amplitudes at Cz were -1.8 and -1.9/zV, in the first and second recording sessions, respectively, their corresponding latencies being 99 and 100 ms. The N1 wave was larger at the central than at the frontal electrodes, and it also showed a contral-
B
A
F3
Fz
.
F4
T
F3
T MMN Fz I
F4
C3
I ~
c~
3uV
"~'~' I1 7i1 ; I- li--¢
standard ist. deviant ist. standard 2nd. deviant 2nd.
Ores
DIFFERENCE WAVE 1st. session 2nd. session
Fig. 2. Grand-averageERPs to standard and deviant tones, and grand-meandifference waves in the two recording sessions.
C. Escera C. Grau / Electroencephalography and clinical Neurophysiology 100 (1996) 549-554
MMN -3.0
-2.5
N1
2.0
oJ
• ..........
o,,
-15
/ ,,<
0.5
v o..
v
1.0
1.5
P2 2.0
L
S
R frontal
1st.
central
1st.
frontal
2nd.
[] ........ o . . . . . . . . o c e n t r a l
2nd.
Fig. 3. Mean peak amplitude distribution for the NI and P2 in the ERPs to standard tones, and for the MMN peak in the difference wave, in the 1st and 2nd recording sessions. L, left electrodes (F3, C3); S, sagittal electrodes (Fz, Cz): R, right electrodes (F4, C4).
the two recording sessions (F(I,10)= 1.08, P = 0.322, n.s.).
3.3. Subjective alertness According to the Stanford Sleepiness Scale, the mean subjects' sleepiness was similar before the first and second recording sessions (scores 2.4 and 2.0, respectively; t(10) = 1.49, P = 0.1669, n.s.). 4. Discussion The results obtained in the present study demonstrated good short-term replicability of the MMN at the group level, and are in agreement with those recently obtained by Pekkonen at al. (1995). These authors found that both the frequency and duration MMNs, obtained in a paradigm similar to that used here, remained stable in a group of 10 healthy young adults as a whole, between two recordings sessions that were performed with an interval of 1 month. The N 1 and P2 components for standard tones presented
553
a fairly good short-term replicability at both the group and individual levels, with no significant differences in either their amplitude or latency between the recording sessions, as revealed by analyses of variance, and by Pearson's product-moment correlation coefficients, which were close to 0.9 at all the electrodes. These results are also in agreement with the well-documented individual and overall replicability of the N1 and P2 in previous studies (Roth et al., 1975; Shelley et al., 199l; Pekkonen et al., 1995). The short-term replicability of the MMN at the individual level was, however, significant at only two electrodes (F4, C3), and approached significance at two additional scalp locations (F3, Fz). This result is consistent with that reported by Pekkonen et al. (1995) who found an individual long-term (1 month) replicability of the MMN at the F4 electrode only, but is in contrast with the high short-term replicability obtained for the N1 and P2 in the ERP to the standard tones in the present study. The lack of a good replicability of the MMN at the individual level reported here is also in agreement with the large intraindividual variation of the MMN amplitude described by Lang et al. (1995), who recorded MMNs from 3 male subjects on 5 different days. by using 4 different deviance conditions. Two reasons might account for this lack of short-term replicability of the MMN at the individual level. First, the involvement of the subjects in the reading task may have varied between the two recording sessions, thus influencing the amplitude of the non-specific component of the MMN. However, since the subject's sleepiness was similar before the two sessions, a possible influence of alertness fluctuations between sessions on MMN amplitude can be ruled out (Lang et al., 1995). On the other hand, the poor short-term replicability of the MMN at the individual level might be explained by the small number of averages for deviant tones. Indeed, in the present study, only 160 deviant and 1840 standard tones were presented in a session. After applying the artifact rejection criteria, at least 130 responses to deviant and 1479 responses to standard tones were averaged to obtain the corresponding ERPs. Moreover, when the individual replicability of the N 1 component studied in the ERP to the deviant tones was analyzed, a significant correlation was found between the two recording sessions only at Fz, but not at Cz. This poor result, as compared with the high individual replicability of the NI in the ERP to the standards, suggests that the number of averages included in the ERP is important to good replicability. By increasing the number of summations, the signal-to-noise ratio increases, and therefore the detection of components in the ERP is made more reliable (Regan, 1989). In conclusion, the present results suggest that the MMN has good short-term replicability at the group level, and therefore, it can be used to study the effectiveness of rehabilitation and treatments in clinical groups. At the individual level, however, the replicability of the MMN should
554
c. Escera, C. Grau / Electroencephalography and clinical Neurophysiology 100 (1996) 549-554
be i m p r o v e d in order to be able to take the m a x i m u m advantage f r o m its p r o m i s i n g clinical applications. This m i g h t be a c h i e v e d , for instance, by increasing the n u m b e r o f deviant tones included in the E R P averages.
Acknowledgements The authors are grateful to J o r g e N o g u e r a and Merc~ Fern~mdez for running the experiments, and to K i m m o A l h o and two a n o n y m o u s referees for v a l u a b l e c o m m e n t s . This study was supported by D G I C Y T grants P B 9 3 - 0 8 0 2 to Caries Escera and P M 9 1 - 0 1 5 2 - C 2 - 0 1 to Caries Grau.
References Alho, K. Cerebral generators of mismatch negativity (MMN) and its magnetic counterpart (MMNn) elicited by sound changes. Ear Hear., 1995, 16: 38-51. Giard, M.H., Perrin, F., Pemier, J. and Bouchet, P. Brain generators implicated in the processing of auditory stimulus deviance: a topographic event-related potential study. Psychophysiology, 1990, 27: 627-640. Hodes, E., Zarcone, V., Smythe, H., Phillips, R. and Dement, W.C. Quantification of sleepiness: a new approach. Psychophysiology, 1973, 10: 431-436. Javitt, D.C., Doneshka, P., Grochowski, S. and Ritter, W. Impaired mismatch negativity generation reflects widespread dysfunction in working memory in schizophrenia. Arch. Gen. Psychiatry, 1995, 52: 550558. Kane, N.M., Curry, S.H., Butler, S.R. and Cummins, B.H. Electrophysiological indicator of awakening from coma. Lancet, 1993, 341: 688. Lang, A.H., Eerola, O., Korpilahti, P., Holopainen, I., Salo, S. and Aal-
tonen, O. Practical issues in the clinical application of the mismatch negativity. Ear Hear., 1995, 16: 118-130. N~tanen, R. Attention and Brain Function, Lawrence Erlbaum, Hillsdale, NJ, 1992. N~t~en, R., Paavilainen, P., Alho, K., Reinikainen, K. and Sams, M. Do event-related potentials reveal the mechanism of the auditory sensory memory in the human brain? Neurosci. Lett., 1989, 98: 217-221. Pekkonen, E., Jousm~ki, V., Partanen, J. and Karhu, J. Mismatch negativity area and age-related auditory memory. Electroenceph. clin. Neurophysiol., 1993, 87: 321-325. Pekkonen, E., Jousm~d, V., K6n6nen, M., Reinikainen, K. and Partanen, J. Auditory sensory memory impairment in Alzheimer's disease: an event-related potential study. NeuroReport, 1994, 5: 2537-2540. Pekkonen, E., Rinne, T. and N~ia't/inen,R. Variability and replicability of the mismatch negativity. Electroenceph. clin. Neurophysiol., 1995, 96: 546-554. Ponton, C.W. and Don, M. The mismatch negativity in cochlear implant users. Ear Hear., 1995, 16: 131-146. Regan, D. Human Brain Electrophysiology, Elsevier, Amsterdam, 1989. Roth, W.T., Kopell, B.S., Tinklenberg, J.R., Huntsberger, G.E. and Kraemer, H.C. Reliability of the contingent negative variation and the auditory evoked potential. Electroenceph. clin. Neurophysiol., 1975, 38: 45-50. Shelley, A.M., Ward, P.B., Michie, P.T., Andrews, S., Mitchell, P.F., Catts, S.V. and McConaghy, N. The effect of repeated testing on ERP components during auditory selective attention. Psychophysiology, 1991, 28: 496-510. Vollmer, D.G., Torner, J.C., Jane, J.A., Sadovnic, B.S., Charlebois, R.N., Eisenberg, H.W., Foulkes, M.A., Marmarou, A. and Marshall, L.F. Age and outcome following traumatic coma: why do older patients fare worse? J. Neurosurg., 1991, 75: 537-549. Woods, D.L. Auditory selective attention in middle-aged and elderly subjects: an event-related brain potential study. Electroenceph. clin. Neurophysiol., 1992, 84: 456-468.
Electroencephalography and clinical Neurophysiology 100 (1996) 549-554
Short-term replicability of the mismatch negativity Caries Escera*,
Caries Grau
Neurod~,namics Laboratory, Department of Psychiatry and Clinical Psychobiology, University c?["Barcehma. P. Vail d'Hebron, 171, 08035 Barcelona. Catalonia. Spain
Accepted for publication: 19 June 1996
Abstract The short-term replicability of the mismatch negativity (MMN) between two recording sessions spaced 2 h apart was evaluated at individual and group levels in a sample of 11 healthy adults. Subjects were presented with a random sequence of 1000 Hz standard (92%) and 1100 Hz deviant (8%) tones while they were reading a book. The N1 and P2 exogenous components to standard tones showed a fairly good individual and group replicability. There were no significant differences in the MMN amplitude and latency between the two sessions in the group of subjects as a whole. The individual replicability of the MMN was not as good as for the NI to standards, reaching significance in only some of the electrodes. This result was, however, similar to that obtained for the NI after deviant tones. The results indicate that the MMN has good replicability at the group level, and further that at the individual level, MMN replicability is similar to that of the N1 to deviants. This suggests that the number of summations should be increased in order to improve the clinical usefulness of the MMN. Keywords: Mismatch negativity (MMN); Sensory memory; Event-related potentials; N1; Auditory function; Test-retest replicability
1. Introduction Infrequent, physically deviant sounds occurring in a series of unattended standard auditory stimuli elicit a component of the event-related potential (ERP) called mismatch negativity (MMN), peaking at 100-250 ms from stimulus onset. The M M N can be obtained by changing any physical attribute of a tone, as well as by changes in more complex sounds (for a review see N~i~it~inen, 1992). Since no M M N can be obtained by presenting the deviant stimulus alone, nor by using long inter-stimulus intervals (ISis), it has been proposed that the physical features of the auditory stimuli are fully analyzed and encoded in neural traces of echoic memory, the M M N being automatically elicited each time the afferent input caused by an auditory stimulus does not match the neuronal representation of the standard stimuli (N~i~it~inen et al., 1989; N~i~it~inen, 1992). Neural generators of M M N have been located in the supratemporal auditory cortex, as revealed by ERPs and magnetic responses, as well as by intracranial recordings * Corresponding author. Tel.: +34 3 4021080, ext. 3047; fax: +34 3 4021584; e-mail: [email protected]
in cats, monkeys and human subjects (reviewed by Alho, 1995). A further contribution to M M N has been located in the right frontal cortex, as revealed by current source density analysis methods (Giard et al., 1990). This frontal activity involved in M M N generation might be associated with an involuntary switching of attention to any discriminable change in the acoustic stimulation (Giard et al., 1990; N~i~it~inen, 1992). The special interest of M M N for clinical applications is that it is elicited independently of the direction of attention, thus making it possible to study auditory discrimination, sensory memory, and involuntary attention in individuals unable or unwilling to cooperate. Among the most promising developments for MMN, the assessment of cognitive brain dysfunction in normal aging (Woods, 1992; Pekkonen et al., 1993), Alzheimer's disease (Pekkonen et al., 1994), and schizophrenia (Javitt et al., 1995), stands out. In addition, M M N may be used to evaluate hearing rehabilitation after cochlear implants (Ponton and Don, 1995), and as a tool in the prognosis of comatose patients (Kane et al., 1993). However, the final clinical utility of M M N will depend on its individual and overall replicability in short-term measures, when diagnostic,
0168-5597/96/$15.00 © 1996 Elsevier Science Ireland Ltd. All rights reserved Pll S0921-884X(96)95633-7
EEP95633
550
C. Escera, C. Grau / Electroencephalography and clinical Neurophysiology 100 (1996) 549-554
prognostic, or follow-up applications are to be taken into account. In the present paper, the short-term replicability and stability of the MMN, at both individual and group levels, will be studied in a group of healthy adults, by comparing the recordings from two sessions spaced 2 h apart. As one of the most promising clinical uses of MMN may be the monitoring of coma state, in order to identify the earliest sign of recovering of consciousness (Kane et al., 1993), the study was designed to fulfill some of its possible requirements. Thus, the recordings were conducted as they may be performed in the intensive care unit, that is, while the subjects were wearing the electrodes in the interviewing period between recording sessions. Further, the mean age of the subjects studied was 35 years, being in the middle of the age-range covering the 89% of the potential population, according to the Traumatic Coma Data Bank report (Vollmer et al., 1991), to whom a possible MMN test for monitoring of coma states might be applied. 2. Methods 2.1. Subjects
Eleven healthy volunteers (4 females), between the ages of 28 and 41 years (mean age 34.5 + 4.0 years) participated in the experiment. None of them had a previous history of neurological or psychiatric problems, nor of alcohol or drug abuse and, unaware of the purpose of the experiment, they gave informed consent to their participation. Subjects were screened in an audiometric test before starting the recording sessions, and all had a threshold below 40 dB SPL. 2.2. Stimuli and procedure
The Stim software and hardware (Neuroscan, Inc., USA) were used to generate pure sine-wave tones of 60 ms, including 5 ms of rise and fall times, which were presented monaurally to the subject's right ear at an intensity of 85 dB SPL through TDH-39 headphones, with a constant ISI (onset-to-onset) of 480 ms. Standard tones were of 1000 Hz in frequency, and were randomly replaced by deviant tones of 1100 Hz with a probability of 0.08. Four different blocks of 500 stimuli were presented in each of the two recording sessions spaced 2 h apart. Each session lasted for about 20 min, the first starting at approximately 1000 h. Subjects sat in a comfortable chair in an electrically shielded and sound-attenuated room, which was dimly lit (340 lux). The subject was instructed to ignore the stimulation and to concentrate on reading a self-selected book. After the first session, the subject was asked to stay in the same room and to sit on the chair, and to try to remain as relaxed as possible. To control possible effects of fluctuations in subject's sleepiness during the
experiment, subjective alertness was evaluated before each recording session by means of a Catalan version of the Stanford Sleepiness Scale (SSS) (Hodes et al., 1973). The EEG (bandpass 0.15-100 Hz) was continuously amplified and digitized at a rate of 250 Hz/channel, from 7 tin scalp electrodes inserted in a nylon cap (ElectroCap, Inc.), according to the 10-20 system: F3, Fz, F4, C3, Cz, C4, and Pz. One supraorbital electrode and another electrode at the canthus of the left eye served to monitor EOG changes. The reference electrode was attached at the tip of nose, and the subject was grounded with an electrode placed between the Fz and Fpz locations. EEG epochs of 390 ms, beginning 40 ms before each stimulus onset, were obtained off-line by a computer, and averaged separately for the standard and deviant tones. Trials exceeding +80 /~V were automatically excluded from the averages, as well as trials containing excessive eye movements, blinks, bursts of muscle activity, amplifier clipping, or other extra-cerebral artifacts. The first 5 trials of each block were also excluded from the averages. After rejection protocols, al least 130 deviant tones in each recording session were averaged for each subject, the mean number of deviant tones averaged being 148 in both sessions. The ERP waveforms were digitally lowpass filtered at 30 Hz. 2.3. Data analysis
The exogenous NI was defined as the most negative peak in the latency window 80-150 ms after stimulus onset, and the P2 as the most positive peak in the 140220 ms window. The MMN was evaluated as the most negative peak in the 130-280 ms latency window in the difference wave obtained by subtracting the ERP to standard tones from that to deviant tones. All amplitude measurements were made against the mean amplitude of the 40 ms pre-stimulus baseline. Two statistical analyses were performed upon the data. First, to test the individual replicability of the MMN, N1, and P2 between the two recording sessions, Pearson's product-moment correlation coefficients (r) between the two sessions were calculated separately for the measurements obtained at three frontal (F3, Fz, F4) and three central (C3, Cz, C4) electrodes. Then, an analysis of variance for repeated measures (electrode x recording session; ANOVA with P values reported after the GreenhouseGeisser correction when appropriate) were applied to study the test-retest replicability at the group level. 3. Results 3.1. Individual replicability 3.1.1. Mismatch negativity Fig. 1 shows the individual difference waves (deviant ERP minus standard ERP) at F4 in the two recording ses-
C. Escera. C. Grau / Electroencephalography and clinical Neurophysiology 100 (1996) 549-554
551
Table 1 Interindividual range and mean peak amplitude ( + SD) in #V and lalency in ms of the MMN at Fz in the two recording sessions Session
1 2
MMN peak (#V)
MMN latency (ms)
Mean _+SD
Largest
Smallest
Mean + SD
Earliest
Latest
-2.6 _+ 1.3 -2.8 _+ 1.4
-5.0 -5.4
-0.6 -0.9
192 + 26.3 195 + 38.0
160 140
228 260
sions for the 11 subjects studied. A large interindividual variation can be seen in the M M N traces. M e a n M M N amplitude and latency, as well as interindividual range, are displayed for the two sessions in Table 1. The individual short-term replicability of the M M N was estimated by the P e a r s o n ' s p r o d u c t - m o m e n t correlation coefficient b e t w e e n the first and second recording sessions, separately for 6 scalp site recordings. Significant correlation coefficients were found only at F4 (r = 0.598, P = 0.04) and C3 (r = 0.657, P = 0.02) for the M M N amplitude. A tendency to a significant short-term replicability o f the M M N ampli-
tude was also found at F3 (r = 0.532, P = 0.075, n.s.) and Fz (r = 0.516, P = 0.086, n.s.). For the M M N latency, the correlation coefficient was significant only at Fz (r = 0.582, P = 0.047). In an additional analysis, after r e m o v i n g the two subjects (d and j) in w h i c h the identification of the M M N peak was less clear (Fig. 1, bottom), the correlation coefficients for M M N amplitude and latency did not reach significance at any electrode. 3.1.2. N I to d e v i a n t t o n e s The same procedure was also used to test the individual
e
I¢t~
,
1
I
+
+
f
J
-5uV Z
k - -
ist. 2nd.
" rV
'
F
V~v '~
v' 5uV
session session
60ms
I
h
i ,~.i
lh
- -
standard
- -
devian[
i
L
Fig. 1. Top: individual difference waves (ERP to deviant tones minus ERP to standard tones) at F4, where MMN amplitude correlation was significant, for the 11 subjects studied (letters from a to k), in the first and second recording sessions. The first session started at 1000 h, and the second session 2 h later. Bottom: examples of individual ERPs to standard and deviant tones (F4 recording, first session) in which subtraction yielded clear (subject h) or less clear (subject d) MMN peak identification (dark area, MMN).
552
C. Escera, C. Grau / Electroencephalography and clinical Neurophysiology 100 (1996) 549-554
ateral (left-hemisphere) dominance (Fig. 3). The two-way analysis of variance did not reveal any significant differences between the recording sessions for either the N1 amplitude (F(1,10)= 0.76, P = 0.403, n.s.) or latency (F(1,10) = 0.97, P = 0.349, n.s.). The P2 deflection had a mean peak amplitude at Cz of 1.2 #V in the first recording session and 1.0 #V in the second, with respective latencies being 168 ms and 165 ms. The P2 was largest at the central electrodes, and at the midline positions (Fig. 3). No significant differences between the recording sessions were found for either the P2 amplitude (F(1,10) = 2.28, P = 0.162, n.s.) or latency (F(1,10) = 0.21, P = 0.660, n.s.).
short-term replicability of the N1 to deviants at Fz and Cz, where the NI amplitude was larger. A significant correlation was found between the N1 amplitude to deviants in the two recording sessions at Fz (r = 0.661, P = 0.019), but not at Cz (r = 0.499, P = 0.098, n.s.). The latency of the N1 to deviants obtained in the two recording sessions did not correlate significantly at any of the electrodes analyzed (r = 0.267, P = 0.412, n.s., at Fz; r = 0.234, P = 0.464, n.s., at Cz). 3.1.3. ERPs to standard tones
Pearson's product-moment correlation coefficient between the first and the second recording sessions at 6 scalp locations were all highly significant for the N1 and P2 amplitudes (ranging from 0.791, P < 0.01, to 0.948, P < 0.001). For the N1 and P2 latencies, the correlation coefficients were significant at all 6 electrodes, except at F3 for N1 latency, and Cz and C4 for P2 latency (ranging from 0.031, n.s., to 0.939, P < 0.001).
3.2.2. Overall replicability o f the M M N
Fig. 2 also shows the grand-mean difference waves obtained by subtracting the ERP to standard tones from those to deviant tones, for the two recording sessions. The scalp distribution of the MMN peak amplitude was different from that of the N1 deflection, with largest amplitudes at frontal electrodes, whereas the N1 was larger at central scalp sites (Fig. 3). The MMN also tended to be larger over the right hemisphere, as opposed to the N1 deflection, which was larger over the left hemisphere (Fig. 3). The MMN peak showed a slight enhancement in its amplitude in the second recording session (Fig. 3). However, a twoway analysis of variance performed with electrode and recording session as factors failed to reveal statistical significance for the factor session (F(1,10) = 0.73, P = 0.412, n.s.), thus showing a high short-term replicability at the group level. The MMN peak latency was also similar in
3.2. Overall replicability 3.2.1. ERPs to standard tones
Grand-average ERPs to standard and deviant tones in the first and second recording sessions are shown in Fig. 2. Standard tones elicited prominent N1 and P2 deflections in both recording sessions. N1 mean peak amplitudes at Cz were -1.8 and -1.9/zV, in the first and second recording sessions, respectively, their corresponding latencies being 99 and 100 ms. The N1 wave was larger at the central than at the frontal electrodes, and it also showed a contral-
B
A
F3
Fz
.
F4
T
F3
T MMN Fz I
F4
C3
I ~
c~
3uV
"~'~' I1 7i1 ; I- li--¢
standard ist. deviant ist. standard 2nd. deviant 2nd.
Ores
DIFFERENCE WAVE 1st. session 2nd. session
Fig. 2. Grand-averageERPs to standard and deviant tones, and grand-meandifference waves in the two recording sessions.
C. Escera C. Grau / Electroencephalography and clinical Neurophysiology 100 (1996) 549-554
MMN -3.0
-2.5
N1
2.0
oJ
• ..........
o,,
-15
/ ,,<
0.5
v o..
v
1.0
1.5
P2 2.0
L
S
R frontal
1st.
central
1st.
frontal
2nd.
[] ........ o . . . . . . . . o c e n t r a l
2nd.
Fig. 3. Mean peak amplitude distribution for the NI and P2 in the ERPs to standard tones, and for the MMN peak in the difference wave, in the 1st and 2nd recording sessions. L, left electrodes (F3, C3); S, sagittal electrodes (Fz, Cz): R, right electrodes (F4, C4).
the two recording sessions (F(I,10)= 1.08, P = 0.322, n.s.).
3.3. Subjective alertness According to the Stanford Sleepiness Scale, the mean subjects' sleepiness was similar before the first and second recording sessions (scores 2.4 and 2.0, respectively; t(10) = 1.49, P = 0.1669, n.s.). 4. Discussion The results obtained in the present study demonstrated good short-term replicability of the MMN at the group level, and are in agreement with those recently obtained by Pekkonen at al. (1995). These authors found that both the frequency and duration MMNs, obtained in a paradigm similar to that used here, remained stable in a group of 10 healthy young adults as a whole, between two recordings sessions that were performed with an interval of 1 month. The N 1 and P2 components for standard tones presented
553
a fairly good short-term replicability at both the group and individual levels, with no significant differences in either their amplitude or latency between the recording sessions, as revealed by analyses of variance, and by Pearson's product-moment correlation coefficients, which were close to 0.9 at all the electrodes. These results are also in agreement with the well-documented individual and overall replicability of the N1 and P2 in previous studies (Roth et al., 1975; Shelley et al., 199l; Pekkonen et al., 1995). The short-term replicability of the MMN at the individual level was, however, significant at only two electrodes (F4, C3), and approached significance at two additional scalp locations (F3, Fz). This result is consistent with that reported by Pekkonen et al. (1995) who found an individual long-term (1 month) replicability of the MMN at the F4 electrode only, but is in contrast with the high short-term replicability obtained for the N1 and P2 in the ERP to the standard tones in the present study. The lack of a good replicability of the MMN at the individual level reported here is also in agreement with the large intraindividual variation of the MMN amplitude described by Lang et al. (1995), who recorded MMNs from 3 male subjects on 5 different days. by using 4 different deviance conditions. Two reasons might account for this lack of short-term replicability of the MMN at the individual level. First, the involvement of the subjects in the reading task may have varied between the two recording sessions, thus influencing the amplitude of the non-specific component of the MMN. However, since the subject's sleepiness was similar before the two sessions, a possible influence of alertness fluctuations between sessions on MMN amplitude can be ruled out (Lang et al., 1995). On the other hand, the poor short-term replicability of the MMN at the individual level might be explained by the small number of averages for deviant tones. Indeed, in the present study, only 160 deviant and 1840 standard tones were presented in a session. After applying the artifact rejection criteria, at least 130 responses to deviant and 1479 responses to standard tones were averaged to obtain the corresponding ERPs. Moreover, when the individual replicability of the N 1 component studied in the ERP to the deviant tones was analyzed, a significant correlation was found between the two recording sessions only at Fz, but not at Cz. This poor result, as compared with the high individual replicability of the NI in the ERP to the standards, suggests that the number of averages included in the ERP is important to good replicability. By increasing the number of summations, the signal-to-noise ratio increases, and therefore the detection of components in the ERP is made more reliable (Regan, 1989). In conclusion, the present results suggest that the MMN has good short-term replicability at the group level, and therefore, it can be used to study the effectiveness of rehabilitation and treatments in clinical groups. At the individual level, however, the replicability of the MMN should
554
c. Escera, C. Grau / Electroencephalography and clinical Neurophysiology 100 (1996) 549-554
be i m p r o v e d in order to be able to take the m a x i m u m advantage f r o m its p r o m i s i n g clinical applications. This m i g h t be a c h i e v e d , for instance, by increasing the n u m b e r o f deviant tones included in the E R P averages.
Acknowledgements The authors are grateful to J o r g e N o g u e r a and Merc~ Fern~mdez for running the experiments, and to K i m m o A l h o and two a n o n y m o u s referees for v a l u a b l e c o m m e n t s . This study was supported by D G I C Y T grants P B 9 3 - 0 8 0 2 to Caries Escera and P M 9 1 - 0 1 5 2 - C 2 - 0 1 to Caries Grau.
References Alho, K. Cerebral generators of mismatch negativity (MMN) and its magnetic counterpart (MMNn) elicited by sound changes. Ear Hear., 1995, 16: 38-51. Giard, M.H., Perrin, F., Pemier, J. and Bouchet, P. Brain generators implicated in the processing of auditory stimulus deviance: a topographic event-related potential study. Psychophysiology, 1990, 27: 627-640. Hodes, E., Zarcone, V., Smythe, H., Phillips, R. and Dement, W.C. Quantification of sleepiness: a new approach. Psychophysiology, 1973, 10: 431-436. Javitt, D.C., Doneshka, P., Grochowski, S. and Ritter, W. Impaired mismatch negativity generation reflects widespread dysfunction in working memory in schizophrenia. Arch. Gen. Psychiatry, 1995, 52: 550558. Kane, N.M., Curry, S.H., Butler, S.R. and Cummins, B.H. Electrophysiological indicator of awakening from coma. Lancet, 1993, 341: 688. Lang, A.H., Eerola, O., Korpilahti, P., Holopainen, I., Salo, S. and Aal-
tonen, O. Practical issues in the clinical application of the mismatch negativity. Ear Hear., 1995, 16: 118-130. N~tanen, R. Attention and Brain Function, Lawrence Erlbaum, Hillsdale, NJ, 1992. N~t~en, R., Paavilainen, P., Alho, K., Reinikainen, K. and Sams, M. Do event-related potentials reveal the mechanism of the auditory sensory memory in the human brain? Neurosci. Lett., 1989, 98: 217-221. Pekkonen, E., Jousm~ki, V., Partanen, J. and Karhu, J. Mismatch negativity area and age-related auditory memory. Electroenceph. clin. Neurophysiol., 1993, 87: 321-325. Pekkonen, E., Jousm~d, V., K6n6nen, M., Reinikainen, K. and Partanen, J. Auditory sensory memory impairment in Alzheimer's disease: an event-related potential study. NeuroReport, 1994, 5: 2537-2540. Pekkonen, E., Rinne, T. and N~ia't/inen,R. Variability and replicability of the mismatch negativity. Electroenceph. clin. Neurophysiol., 1995, 96: 546-554. Ponton, C.W. and Don, M. The mismatch negativity in cochlear implant users. Ear Hear., 1995, 16: 131-146. Regan, D. Human Brain Electrophysiology, Elsevier, Amsterdam, 1989. Roth, W.T., Kopell, B.S., Tinklenberg, J.R., Huntsberger, G.E. and Kraemer, H.C. Reliability of the contingent negative variation and the auditory evoked potential. Electroenceph. clin. Neurophysiol., 1975, 38: 45-50. Shelley, A.M., Ward, P.B., Michie, P.T., Andrews, S., Mitchell, P.F., Catts, S.V. and McConaghy, N. The effect of repeated testing on ERP components during auditory selective attention. Psychophysiology, 1991, 28: 496-510. Vollmer, D.G., Torner, J.C., Jane, J.A., Sadovnic, B.S., Charlebois, R.N., Eisenberg, H.W., Foulkes, M.A., Marmarou, A. and Marshall, L.F. Age and outcome following traumatic coma: why do older patients fare worse? J. Neurosurg., 1991, 75: 537-549. Woods, D.L. Auditory selective attention in middle-aged and elderly subjects: an event-related brain potential study. Electroenceph. clin. Neurophysiol., 1992, 84: 456-468.