- Email: [email protected]
An alternative baseline measure for calculation of the mismatch negativity (MMN) component of the event-related potential (ERP)
International Journal of Psychophysiology 51 (2004) 201–208
An alternative baseline measure for calculation of the mismatch negativity (MMN) component of the event-related potential (ERP) Gerene K. Starratta,*, Allan J. Nashb a
Department of Psychology, Barry University, 11300 Northeast Second Avenue, Miami Shores, FL 33161-6695, USA b Department of Psychology, Florida Atlantic University, P.O. Box 3091, Boca Raton, FL 33431-0991, USA Received 14 January 2003; received in revised form 16 September 2003; accepted 23 September 2003
Abstract Mismatch negativity (MMN), a component of the event-related potential, reflects the processing associated with low-probability (deviant) stimuli over and above that required for high-probability (standard) stimuli. Occurring 100– 250 ms post stimulus, MMN is evident in the enhanced negativity generated in response to deviant relative to standard stimuli. Traditionally, the MMN waveform is calculated by subtracting the averaged waveform of all standard stimuli from the averaged waveform of all deviant stimuli collected during the same test session. To investigate whether an unbroken, extended string of standard stimuli may minimize baseline processing, thereby resulting in a more robust measure of MMN, 11 participants were exposed to two contiguous blocks of 1000 trials, each beginning with 60 uninterrupted standard stimuli followed by 940 randomized standard and deviant trials. Negativity related to standard stimuli was minimized in the average of uninterrupted standard stimuli from the second block of trials only while negativity related to uninterrupted standards in the first block of trials was increased relative to all other conditions. The present findings suggest that a familiar, uninterrupted string of standard stimuli, presented midway through an experimental session, may provide a useful alternative baseline for the calculation of MMN. 䊚 2003 Elsevier B.V. All rights reserved. Keywords: Mismatch negativity; Event-related potentials; Methodology; Electrophysiology; Psychophysiology
1. Introduction Mismatch negativity (MMN) is a component of the event-related potential (ERP) that was first ¨¨ ¨ reported by Naatanen (1979). Characterized as a change-detector response, this component purportedly provides an indirect physiological measure of the brain’s processing of sensory information. *Corresponding author. Tel.: q1-305-899-4575; fax: q1305-899-3279. E-mail address: [email protected] (G.K. Starratt).
Cheour-Luhtanen et al. (1996) and Cheour et al. (2000), who have identified a MMN-like component in premature infants, suggest that MMN constitutes the ontogenetically earliest discriminative response that has been measured in humans. MMN is generally elicited in an oddball paradigm in which the participant is presented with a string of auditory stimuli consisting of only 2 tones that differ on some characteristic (e.g. frequency, intensity, etc.). One of the tones (the standard) occurs at high probability (e.g. at least
0167-8760/04/$ - see front matter 䊚 2003 Elsevier B.V. All rights reserved. doi:10.1016/j.ijpsycho.2003.09.008
202
G.K. Starratt, A.J. Nash / International Journal of Psychophysiology 51 (2004) 201–208
80% of all trials) and the target tone, in contrast, occurs at low-probability. The MMN component, ¨¨ ¨ as originally defined by Naatanen, is evident in the ERP waveform as an enhanced negativity that occurs approximately 100–250 ms following stimulus onset in response to the deviant stimulus as ¨¨ ¨ compared to the standard stimulus (Naatanen, 1992). This component can be identified at frontal, central and parietal electrode locations, but is generally attenuated at parietal sites in relation to frontal and central sites. The fundamental characteristic of MMN is that it is evident only in response to changes in stimuli. MMN is reportedly not elicited to an initial stimulus or under intersti¨¨ ¨ mulus intervals (ISIs) of 4 s or longer (Naatanen, 1992). Because MMN manifests as an increased negativity in the deviant stimulus waveform relative to the standard stimulus waveform, it is generally depicted as a difference waveform that is obtained by subtracting the averaged ERP waveform of all standard stimuli from the averaged ERP waveform associated with all deviant stimuli presented in the same oddball condition. Consequently, the MMN difference waveform purportedly reflects only the negativity in processing related to the deviant stimulus over and above the processing devoted to ¨¨ ¨ the standard stimulus. Naatanen (1992) has noted that general increases in activation are associated with increased MMN and general decreases in activation are associated with decreased MMN. Of relevance to the present project, it has been reported that MMN is also elicited to non-deviant stimuli under certain conditions. For example, Sams et al. (1984) reported MMN to standard stimuli that immediately follow a deviant stimulus. Similarly, it has been reported that in a string of equiprobable stimuli (i.e. 50–50 probability, stimulus A and stimulus B), MMN is elicited to either stimulus when it is preceded by a string of several exemplars of the other stimulus (Sams et al., 1983). Similar results were reported by Winkler et al. (1990) who, in presenting multiple standard and multiple deviant stimuli, reported that MMN was generated to all standard and deviant stimuli. Furthermore, these authors also reported that when the pitch of the standard tone was varied slightly, resulting in multiple similar ‘substandard’ tones,
MMN was elicited to even the most similar substandard tones and increased in amplitude with increasing deviance from the standard tone. The purpose of the present project was to investigate whether brain responses to standard stimuli might be minimized in the absence of deviant stimuli. Specifically, it was proposed that an uninterrupted string of standard stimuli would yield an ERP waveform reflecting a minimum amount of negativity in the MMN window. MMN theory predicts that this standard waveform, based on an uninterrupted sequence, should show no evidence of the negativity associated with MMN because there is no deviation from the standard tone and, thus, no mismatch to register. Indeed, Sams et al. (1984) have reported that standards presented in an uninterrupted sequence do not elicit any evidence of MMN. It was further predicted that this waveform (theoretically devoid of MMN), when subtracted from the average waveform associated with deviant stimuli obtained in an oddball condition during the same test session, would result in a MMN waveform that would reflect increased MMN-related negativity in comparison to the MMN difference waveform that is calculated using the traditional method. Because data in the present study were gathered in two contiguous blocks, it was also possible to examine the possibility of block differences in ERP responses to uninterrupted standards. 2. Materials and methods 2.1. Subjects Eleven graduate and undergraduate students (eight females and three males, ages 18–26) were participants in this experiment. All participants were volunteers and all reported normal hearing. Some undergraduate participants received extra credit in psychology courses for their participation. All participants provided informed consent. 2.2. EEG recording Scalp recordings were made with silver–silver chloride cup electrodes at midline frontal, vertex and parietal scalp locations in accordance with the
G.K. Starratt, A.J. Nash / International Journal of Psychophysiology 51 (2004) 201–208
International 10–20 System (Jasper, 1958). Additional electrodes were attached as follows: one above and one at the outer canthus of the right eye in order to monitor eye movement artifact, one to each earlobe to act as a linked reference and a plate electrode was attached to the left forearm to act as a ground. Impedances for the active electrodes were maintained under 5 kV. The EEG and EOG signals were amplified and recorded on paper by a Grass 79D polygraph with 7P511 amplifiers that were set at a gain of 25 000 for the active EEG electrode sites. Upper and lower band pass limits were set at 0.1 and 100 Hz, respectively. Trials were rejected when the EOG exceeded 50 mV. Additionally, trials contaminated by alpha, excessive eye-movement or other muscle-related artifact were rejected by visual inspection of the records following the test session. A 996-ms epoch was digitized on a trial-by-trial basis with a sampling rate of one data point approximately every 2 ms, to yield a total of 509 points per digitized epoch. A pre-stimulus interval of 120 ms was included in the digitizing procedure. 2.3. Stimuli Auditory stimuli (sinusoidal tone pulses delivered with a 40-ms onset ramp, 40-ms sustained amplitude and exponential decay) were generated by a Polyfusion Series 2000 Synthesizer. The auditory tones were presented binaurally via shielded headphones at an intensity of 74 dB SPL (re: 20 UPa). Standard and deviant tones were 1000 and 500 Hz, respectively. Stimuli were presented at a fixed 2.5-s ISI. Two identical blocks of 1000 trials were presented in one 90-min session interrupted midway by a 5-min break.
203
slightly more than 2 m from the participant. A 5min break was introduced between the two sessions and, additionally, each session was interrupted halfway through in order to permit the participant to move about for comfort. The movie soundtrack was set to a level where the dialogue could be comfortably understood and the experimental tones were distinctly audible above the soundtrack of the movie. 2.5. Experimental conditions 2.5.1. Baseline condition In an attempt to establish a waveform containing a minimum of negativity in the MMN range, the participant was presented with an uninterrupted string of identical (ns60) standard tones at the beginning of each of the two blocks of trials. Because the first stimulus presentation purportedly results in a greatly inflated N1 response (e.g. Pekkonen et al., 1996), the ERP gathered to the first 10 stimulus presentations was not included in the average of the baseline waveform. The remaining 50 trials of standard stimuli, which comprised the baseline condition data, were coded separately for each block of trials. 2.5.2. Oddball condition The first 60 standard tones of each block of trials were followed, without pause, by 940 standard and deviant tones in a conventional oddball paradigm (f80y20 probability). In this condition, ERPs were collected for both deviant and standard tone trials. Because the first deviant tone in each block of trials followed 60 standard tones, these two deviant stimuli were not included in any analysis. Oddball condition data were not analyzed separately by block.
2.4. Procedure 2.6. Statistical analysis During the testing session, the participant was seated in a comfortable chair in a dimly illuminated sound-attenuating chamber. ERP data were collected while participants viewed a self-selected full-length feature film over approximately a 90min period. To minimize eye movement, the video was presented at eye level on a 13-inch television screen encased in a darkened box at a distance of
MMN was first calculated in the traditional way by subtracting the average waveform of all standards presented in the oddball condition from the average waveform of all deviant stimuli. For the purposes of comparison, MMN was also calculated by subtracting the average waveform of uninterrupted standard stimuli from the average of all
204
G.K. Starratt, A.J. Nash / International Journal of Psychophysiology 51 (2004) 201–208
Fig. 1. Grand average waveforms recorded in response to uninterrupted baseline standards (by block), oddball standards and deviant stimuli, recorded at frontal (Fz) electrode. Arrow indicates stimulus onset.
deviant stimuli. Because MMN is characteristically ¨¨ ¨ defined as occurring at frontal locations (Naatanen, 1992), only data collected at electrode location Fz was analyzed. To maintain alpha at Ps0.05 as the criterion for declaring an effect statistically significant, for each set of comparisons the Bonferroni correction for multiple contrasts was used (i.e. 0.05yc, where csthe number of contrasts in the set, Kirk, 1982). All graphs of ERP waveforms depict a 420-ms epoch that includes a 120-ms prestimulus baseline and 300 ms of data following the presentation of the auditory stimulus. Each waveform has been baseline corrected by subtracting the average of the prestimulus baseline points from all data points in the ERP waveform. In all figures, the arrow indicates stimulus onset and the vertical ticks represent 100, 200 and 300 ms following presentation of the auditory stimulus. For each comparison, analyses were conducted on the sum of ERP or MMN data points across 25-ms windows within the definitional time frame of MMN. These windows were centered on 100,
125, 150, 175, 200 and 225 ms following stimulus onset. In all figures, statistically significant and non-significant effects are illustrated by black and gray bars, respectively. 3. Results The grand averages of the ERPs collected in response to standard stimuli presented in the baseline condition (blocks 1 and 2), all standard stimuli presented in the oddball condition, as well as all deviant stimuli presented in the oddball condition are presented in Fig. 1. As is customarily observed, the waveform that depicts the grand average of all ERP responses to deviant stimuli shows an enhanced negativity in the MMN window as compared to the waveform associated with the average of all standards presented in the oddball condition (oddball standards). Of relevance to the present study, the expected minimum negativity was, indeed, observed in the baseline condition presented at the beginning of the second block of trials (standards block 2). However, it
G.K. Starratt, A.J. Nash / International Journal of Psychophysiology 51 (2004) 201–208
should be noted that the waveform representing baseline standards presented at the beginning of the first block of trials (standards block 1) showed an unexpected increase in negativity in the MMN window as compared to all other stimuli. In fact, the phenomenon of interest in Fig. 1 is the relationship between the waveforms elicited by the three classes of standard stimuli. Fig. 2 depicts only the waveforms associated with uninterrupted standard stimuli presented in the baseline condition in blocks 1 and 2. The ERP response to the uninterrupted standards presented in the first block of trials (i.e. the first stimuli to which participants were exposed) shows a clear increased negativity in the MMN window as compared to uninterrupted standards presented at the beginning of the second block of trials. Analysis of the baseline corrected sum of ERP points for six consecutive 25-ms windows from 88 to 237 ms reveals significantly greater negativity for baseline standards block 1 than for block 2 from 113 to 212 ms following stimulus onset. Means, S.D., F and P values can be found in Table 1.
205
Table 1 Mean (S.D.), F and P values for the sum of data points across successive 25-ms windows for baseline standard stimuli by block Block 1 Block 2 Baseline standards Baseline standards 88–112 ms windowa y12.33 (22.21) 113–137 ms windowb y21.40 (21.32) 138–162 ms windowc y8.73 (15.84) 163–187 ms windowd y0.12 (17.91) 188–212 ms windowe 4.79 (19.81) 213–237 ms windowf 8.18 (20.86)
y3.07 y1.69 25.07 37.09 32.91 24.21
(20.13) (21.58) (18.10) (18.24) (14.99) (13.24)
Bonferroni Ps0.0008. F(1, 10)s4.62, Ps0.06. b F(1, 10)s49.84, *P-0.0001. c F(1, 10)s40.92, *Ps0.0001. d F(1, 10)s33.20, *Ps0.0002. e F(1, 10)s27.52, *Ps0.0004. f F(1, 10)s5.59, Ps0.04. a
The MMN waveforms in Fig. 3 were calculated from the waveforms shown in Fig. 1. As is commonly done, the traditional MMN waveform in this figure results from the subtraction of the
Fig. 2. Grand average waveforms recorded in response to baseline standards in blocks 1 and 2 (from Fig. 1). Arrow indicates stimulus onset. Data recorded at frontal (Fz) electrode. Intervals showing statistically significant (Bonferroni Ps0.008) and nonsignificant differences between the waveforms are indicated with black and gray bars, respectively.
206
G.K. Starratt, A.J. Nash / International Journal of Psychophysiology 51 (2004) 201–208
waveform representing the response to standards in the oddball condition (oddball standards) from the waveform related to deviant stimuli. Similarly, the alternate method MMN waveform depicts a subtraction of the ERP waveform representing responses to baseline standards block 2 from the same deviant waveform. Consequently, this figure reflects a comparison of two methods of calculating MMN. Note that, MMN is more negative in the waveform calculated using the baseline standards block 2. This difference is statistically significant (Bonferroni Ps0.01) from 163 to 237 ms. Means, S.D., F and P values can be found in Table 2.
Table 2 Mean (S.D.), F and P values for the sum of data points across successive 25-ms windows for MMN by method of calculation
4. Discussion
only for the uninterrupted string of standards presented at the beginning of the second block of trials (standards block 2). That is, even though the uninterrupted string of standard stimuli presented to the participant at the beginning of the first and second blocks of trials were identical, these stimuli, when initially presented to the participant, show a significantly enhanced negativity during
The results demonstrate that, as predicted, an uninterrupted string of standard stimuli does elicit reduced negativity in the time frame of MMN (i.e. f100–200 ms following stimulus onset) when compared to the average of all standards collected in the oddball condition. However, this was true
Traditional method Alternate method 113–137 138–162 163–187 188–212 213–237
ms ms ms ms ms
a
window y7.55 (6.48) windowb y12.75 (8.95) windowc y5.08 (12.38) 5.23 (8.93) windowd 6.28 (6.54) windowe
y12.11 y23.64 y15.20 y4.71 y4.16
(16.73) (17.38) (16.55) (11.41) (14.73)
Bonferroni Ps0.01. a F(1, 10)s1.29, Ps0.28. b F(1, 10)s8.17, Ps0.02. c F(1, 10)s11.37, *P-0.01. d F(1, 10)s9.78, *Ps0.01. e F(1, 10)s8.63, *Ps0.01.
Fig. 3. Grand average MMN waveforms calculated using traditional vs. alternate methods. Arrow indicates stimulus onset. Data recorded at frontal (Fz) electrode. Intervals showing statistically significant (Bonferroni Ps0.01) and non-significant differences between the waveforms are indicated with black and gray bars, respectively.
G.K. Starratt, A.J. Nash / International Journal of Psychophysiology 51 (2004) 201–208
this time frame (as compared to all other standard and deviant stimulus conditions), which returns to baseline levels by approximately 250 ms. Conversely, when participants were presented with the same stimuli following the first 1000 trials (and a 5-min break), the result is a greatly attenuated negativity (compared to all other standard and deviant stimulus conditions) in the same time frame. The enhanced negativity associated with the processing of initial standard stimuli in this case may reflect the additional processing resources allotted to novel stimuli by participants who were ¨ to electrophysiological methodology. Connaıve versely, the attenuated negativity associated with processing of the uninterrupted standards at the beginning of the second block of trials may reflect increased comfort and familiarity with the testing situation and the stimuli in general. That is, it is reasonable to assume that the processing of uninterrupted standard stimuli later in the experimental session would reflect some minimum level of processing associated with maintaining vigilance in a familiar environment. Consistent with these baseline differences, MMN calculated by the proposed method (deviant minus standards block 2) showed increased negativity in relation to the traditional method (deviant minus oddball standards). The significant difference between these MMN waveforms occurs between 138 and 237 ms, is in the expected direction and reflects a more theoretically conservative measure of MMN. It has been reported that the signal-to-noise ratio (SNR) in MMN can be problematic (e.g. McGee et al., 2001). The present study suggests that one way to address this problem may be to include a condition that presents an uninterrupted string of identical standard stimuli for the purposes of providing a baseline measure reflecting an absence of MMN. Use of this baseline method yields a measure of MMN that shows increased negativity in the MMN window that does not reflect simply an arbitrary inflation of this measure, but provides a conceptually rigorous index of the MMN phenomenon. It should be noted that the MMN generated by the alternate method not only reflects a larger MMN peak, but also depicts a different MMN
207
morphology. Specifically, in Fig. 2 the alternate method MMN demonstrates an extended period of MMN that lasts almost until 250 ms (i.e. almost twice as long as the traditional MMN). This later epoch observed in the proposed MMN is not seen in the traditional waveform as a result of its presence in both the deviant and oddball standard waveforms. It appears that this difference in morphology may reflect an N2 component that is present in the deviant waveform and in the waveform associated with oddball standards, but not present (or greatly attenuated) in the standards block 2 condition. Although N2 is reportedly elicited to target stimuli (e.g. Alho et al., 1994), if this does indeed represent an N2 component, it appears that there may be some vestige of this component in even unattended stimuli that is present, to an equal degree, in both standard and deviant stimuli in an oddball paradigm. As a design issue, these results indicate that the presence of negativity in the waveform is not influenced only by its immediate context (i.e. the immediately surrounding stimuli), but by its placement in the entire context of the experimental session as well. Specifically, the participants in the present experiment were exposed to two identical 1000-trial blocks, each of which began with 60 uninterrupted standard stimuli. These standard stimuli when presented in the first block, however, elicited increased negativity relative to all other standard and deviant waveforms. Of note, this negativity is increased only in the MMN window as the waveform returns to baseline with the others at approximately 250 ms. Conversely, the significantly attenuated negativity is seen only in the waveform depicting responses to uninterrupted standards when they are presented at the beginning of the second block of trials. From a design standpoint, this divergence indicates that the placement of the uninterrupted standards in the experimental situation is an important issue. In the present study, the minimum negativity was observed in response to a second set of uninterrupted standards presented midway through a 90-min testing session when participants had been informed to expect exactly the same sequence to which they had already been exposed.
208
G.K. Starratt, A.J. Nash / International Journal of Psychophysiology 51 (2004) 201–208
Under what conditions might an absolute minimum MMN be observed? First, it may be that stimulus processing will be minimized in a string of uninterrupted standard stimuli to which there has been previous exposure when the subject has been informed to expect the recurrence. However, if an unannounced, uninterrupted string of standard stimuli occurred unexpectedly in the middle of an oddball paradigm, this would likely be interpreted as a novel change in the experimental situation and would not be expected to reduce processing but, if anything, might even increase negativity associated with processing of the second set of uninterrupted standards. Second, it is possible that presenting the baseline condition even later, perhaps near the end of the session might further reduce the negativity in this waveform. However, it is also reasonable to assume that some sort of expectation effect could manifest near the end of a test session when the participant anticipates that the testing session is drawing to a close. It is likely that this type of anticipation effect would also result in increased rather than decreased negativity. While the experimental conditions for the present study may not be representative of typical MMN studies, which often utilize shorter ISIs and a smaller discrepancy between standard and deviant tones, the current procedures are consistent with conditions that have been previously reported to be appropriate for eliciting a MMN component ¨¨ ¨ (reviewed by Naatanen, 1992). Future studies may address the question of whether these results are replicable under diverse experimental conditions. In conclusion, the present results indicate that a sequence of uninterrupted auditory standard stimuli that is familiar and expected, appears to reduce the negativity associated with processing these stimuli. Consequently, standard stimuli presented under these conditions may provide an alternate baseline measure with reduced error for the calculation of MMN. The ability to improve the SNR for this component yields a theoretically conservative and authentic index of MMN, which can
increase the probability that the results of an experimental manipulation will be detectable. This may permit the design of new experimental methodologies that may ultimately yield new insights into these elusive phenomena and their relationship to fundamental theoretical questions about the nature of human cognition. References Alho, K., Woods, D.L., Algazi, A., 1994. Processing of auditory stimuli during auditory and visual attention as revealed by event-related potentials. Psychophysiology 31, 469–479. Cheour, M., Leppaenen, P.H.T., Kraus, N., 2000. Mismatch negativity (MMN) as a tool for investigating auditory discrimination and sensory memory in infants and children. Clin. Neurophysiol. 111, 4–16. Cheour-Luhtanen, M., Alho, K., Sainio, K., et al., 1996. The ontogenetically earliest discriminative response of the human brain. Psychophysiology 33, 478–481. Jasper, H.H., 1958. The ten–twenty electrode system. Electroencephalogr. Clin. Neurophysiol. 10, 371–375. Kirk, R.E., 1982. Experimental design. second ed.. Brooksy Cole, Monterey, CA. McGee, T.J., King, C., Tremblay, K., Nicole, T.G., Cunningham, J., Kraus, N., 2001. Long-term habituation of the speech-elicited mismatch negativity. Psychophysiology 38, 653–658. ¨¨ ¨ Naatanen, R., 1979. Orienting and evoked potentials. In: Kimmel, H.D., van Olst, E.H., Orlebeke, J.F. (Eds.), The Orienting Reflex in Humans. Lawrence Erlbaum Associates, Hillsdale, NJ, pp. 61–75. ¨¨ ¨ Naatanen, R., 1992. Attention and Brain Function. Lawrence Erlbaum, Hillsdale, NJ. Pekkonen, E., Rinne, T., Reinikainen, K., Kujala, T., Alho, K., ¨¨ ¨ Naatanen, R., 1996. Aging effects on auditory processing: an event-related potential study. Exp. Aging Res. 22, 171–184. ¨¨ ¨ Sams, M., Alho, K., Naatanen, R., 1983. Sequential effects on the ERP in discriminating two stimuli. Biol. Psychol. 17, 41–58. ¨¨ ¨ Sams, M., Alho, K., Naatanen, R., 1984. Short-term habituation and dishabituation of the mismatch negativity of the ERP. Psychophysiology 21, 434–441. Winkler, I., Paavilainen, P., Alho, K., Reinikainen, K., Sams, ¨¨ ¨ M., Naatanen, R., 1990. The effect of small variation of the frequent auditory stimulus on the event-related brain potential to the infrequent stimulus. Psychophysiology 27, 228–235.
An alternative baseline measure for calculation of the mismatch negativity (MMN) component of the event-related potential (ERP) Gerene K. Starratta,*, Allan J. Nashb a
Department of Psychology, Barry University, 11300 Northeast Second Avenue, Miami Shores, FL 33161-6695, USA b Department of Psychology, Florida Atlantic University, P.O. Box 3091, Boca Raton, FL 33431-0991, USA Received 14 January 2003; received in revised form 16 September 2003; accepted 23 September 2003
Abstract Mismatch negativity (MMN), a component of the event-related potential, reflects the processing associated with low-probability (deviant) stimuli over and above that required for high-probability (standard) stimuli. Occurring 100– 250 ms post stimulus, MMN is evident in the enhanced negativity generated in response to deviant relative to standard stimuli. Traditionally, the MMN waveform is calculated by subtracting the averaged waveform of all standard stimuli from the averaged waveform of all deviant stimuli collected during the same test session. To investigate whether an unbroken, extended string of standard stimuli may minimize baseline processing, thereby resulting in a more robust measure of MMN, 11 participants were exposed to two contiguous blocks of 1000 trials, each beginning with 60 uninterrupted standard stimuli followed by 940 randomized standard and deviant trials. Negativity related to standard stimuli was minimized in the average of uninterrupted standard stimuli from the second block of trials only while negativity related to uninterrupted standards in the first block of trials was increased relative to all other conditions. The present findings suggest that a familiar, uninterrupted string of standard stimuli, presented midway through an experimental session, may provide a useful alternative baseline for the calculation of MMN. 䊚 2003 Elsevier B.V. All rights reserved. Keywords: Mismatch negativity; Event-related potentials; Methodology; Electrophysiology; Psychophysiology
1. Introduction Mismatch negativity (MMN) is a component of the event-related potential (ERP) that was first ¨¨ ¨ reported by Naatanen (1979). Characterized as a change-detector response, this component purportedly provides an indirect physiological measure of the brain’s processing of sensory information. *Corresponding author. Tel.: q1-305-899-4575; fax: q1305-899-3279. E-mail address: [email protected] (G.K. Starratt).
Cheour-Luhtanen et al. (1996) and Cheour et al. (2000), who have identified a MMN-like component in premature infants, suggest that MMN constitutes the ontogenetically earliest discriminative response that has been measured in humans. MMN is generally elicited in an oddball paradigm in which the participant is presented with a string of auditory stimuli consisting of only 2 tones that differ on some characteristic (e.g. frequency, intensity, etc.). One of the tones (the standard) occurs at high probability (e.g. at least
0167-8760/04/$ - see front matter 䊚 2003 Elsevier B.V. All rights reserved. doi:10.1016/j.ijpsycho.2003.09.008
202
G.K. Starratt, A.J. Nash / International Journal of Psychophysiology 51 (2004) 201–208
80% of all trials) and the target tone, in contrast, occurs at low-probability. The MMN component, ¨¨ ¨ as originally defined by Naatanen, is evident in the ERP waveform as an enhanced negativity that occurs approximately 100–250 ms following stimulus onset in response to the deviant stimulus as ¨¨ ¨ compared to the standard stimulus (Naatanen, 1992). This component can be identified at frontal, central and parietal electrode locations, but is generally attenuated at parietal sites in relation to frontal and central sites. The fundamental characteristic of MMN is that it is evident only in response to changes in stimuli. MMN is reportedly not elicited to an initial stimulus or under intersti¨¨ ¨ mulus intervals (ISIs) of 4 s or longer (Naatanen, 1992). Because MMN manifests as an increased negativity in the deviant stimulus waveform relative to the standard stimulus waveform, it is generally depicted as a difference waveform that is obtained by subtracting the averaged ERP waveform of all standard stimuli from the averaged ERP waveform associated with all deviant stimuli presented in the same oddball condition. Consequently, the MMN difference waveform purportedly reflects only the negativity in processing related to the deviant stimulus over and above the processing devoted to ¨¨ ¨ the standard stimulus. Naatanen (1992) has noted that general increases in activation are associated with increased MMN and general decreases in activation are associated with decreased MMN. Of relevance to the present project, it has been reported that MMN is also elicited to non-deviant stimuli under certain conditions. For example, Sams et al. (1984) reported MMN to standard stimuli that immediately follow a deviant stimulus. Similarly, it has been reported that in a string of equiprobable stimuli (i.e. 50–50 probability, stimulus A and stimulus B), MMN is elicited to either stimulus when it is preceded by a string of several exemplars of the other stimulus (Sams et al., 1983). Similar results were reported by Winkler et al. (1990) who, in presenting multiple standard and multiple deviant stimuli, reported that MMN was generated to all standard and deviant stimuli. Furthermore, these authors also reported that when the pitch of the standard tone was varied slightly, resulting in multiple similar ‘substandard’ tones,
MMN was elicited to even the most similar substandard tones and increased in amplitude with increasing deviance from the standard tone. The purpose of the present project was to investigate whether brain responses to standard stimuli might be minimized in the absence of deviant stimuli. Specifically, it was proposed that an uninterrupted string of standard stimuli would yield an ERP waveform reflecting a minimum amount of negativity in the MMN window. MMN theory predicts that this standard waveform, based on an uninterrupted sequence, should show no evidence of the negativity associated with MMN because there is no deviation from the standard tone and, thus, no mismatch to register. Indeed, Sams et al. (1984) have reported that standards presented in an uninterrupted sequence do not elicit any evidence of MMN. It was further predicted that this waveform (theoretically devoid of MMN), when subtracted from the average waveform associated with deviant stimuli obtained in an oddball condition during the same test session, would result in a MMN waveform that would reflect increased MMN-related negativity in comparison to the MMN difference waveform that is calculated using the traditional method. Because data in the present study were gathered in two contiguous blocks, it was also possible to examine the possibility of block differences in ERP responses to uninterrupted standards. 2. Materials and methods 2.1. Subjects Eleven graduate and undergraduate students (eight females and three males, ages 18–26) were participants in this experiment. All participants were volunteers and all reported normal hearing. Some undergraduate participants received extra credit in psychology courses for their participation. All participants provided informed consent. 2.2. EEG recording Scalp recordings were made with silver–silver chloride cup electrodes at midline frontal, vertex and parietal scalp locations in accordance with the
G.K. Starratt, A.J. Nash / International Journal of Psychophysiology 51 (2004) 201–208
International 10–20 System (Jasper, 1958). Additional electrodes were attached as follows: one above and one at the outer canthus of the right eye in order to monitor eye movement artifact, one to each earlobe to act as a linked reference and a plate electrode was attached to the left forearm to act as a ground. Impedances for the active electrodes were maintained under 5 kV. The EEG and EOG signals were amplified and recorded on paper by a Grass 79D polygraph with 7P511 amplifiers that were set at a gain of 25 000 for the active EEG electrode sites. Upper and lower band pass limits were set at 0.1 and 100 Hz, respectively. Trials were rejected when the EOG exceeded 50 mV. Additionally, trials contaminated by alpha, excessive eye-movement or other muscle-related artifact were rejected by visual inspection of the records following the test session. A 996-ms epoch was digitized on a trial-by-trial basis with a sampling rate of one data point approximately every 2 ms, to yield a total of 509 points per digitized epoch. A pre-stimulus interval of 120 ms was included in the digitizing procedure. 2.3. Stimuli Auditory stimuli (sinusoidal tone pulses delivered with a 40-ms onset ramp, 40-ms sustained amplitude and exponential decay) were generated by a Polyfusion Series 2000 Synthesizer. The auditory tones were presented binaurally via shielded headphones at an intensity of 74 dB SPL (re: 20 UPa). Standard and deviant tones were 1000 and 500 Hz, respectively. Stimuli were presented at a fixed 2.5-s ISI. Two identical blocks of 1000 trials were presented in one 90-min session interrupted midway by a 5-min break.
203
slightly more than 2 m from the participant. A 5min break was introduced between the two sessions and, additionally, each session was interrupted halfway through in order to permit the participant to move about for comfort. The movie soundtrack was set to a level where the dialogue could be comfortably understood and the experimental tones were distinctly audible above the soundtrack of the movie. 2.5. Experimental conditions 2.5.1. Baseline condition In an attempt to establish a waveform containing a minimum of negativity in the MMN range, the participant was presented with an uninterrupted string of identical (ns60) standard tones at the beginning of each of the two blocks of trials. Because the first stimulus presentation purportedly results in a greatly inflated N1 response (e.g. Pekkonen et al., 1996), the ERP gathered to the first 10 stimulus presentations was not included in the average of the baseline waveform. The remaining 50 trials of standard stimuli, which comprised the baseline condition data, were coded separately for each block of trials. 2.5.2. Oddball condition The first 60 standard tones of each block of trials were followed, without pause, by 940 standard and deviant tones in a conventional oddball paradigm (f80y20 probability). In this condition, ERPs were collected for both deviant and standard tone trials. Because the first deviant tone in each block of trials followed 60 standard tones, these two deviant stimuli were not included in any analysis. Oddball condition data were not analyzed separately by block.
2.4. Procedure 2.6. Statistical analysis During the testing session, the participant was seated in a comfortable chair in a dimly illuminated sound-attenuating chamber. ERP data were collected while participants viewed a self-selected full-length feature film over approximately a 90min period. To minimize eye movement, the video was presented at eye level on a 13-inch television screen encased in a darkened box at a distance of
MMN was first calculated in the traditional way by subtracting the average waveform of all standards presented in the oddball condition from the average waveform of all deviant stimuli. For the purposes of comparison, MMN was also calculated by subtracting the average waveform of uninterrupted standard stimuli from the average of all
204
G.K. Starratt, A.J. Nash / International Journal of Psychophysiology 51 (2004) 201–208
Fig. 1. Grand average waveforms recorded in response to uninterrupted baseline standards (by block), oddball standards and deviant stimuli, recorded at frontal (Fz) electrode. Arrow indicates stimulus onset.
deviant stimuli. Because MMN is characteristically ¨¨ ¨ defined as occurring at frontal locations (Naatanen, 1992), only data collected at electrode location Fz was analyzed. To maintain alpha at Ps0.05 as the criterion for declaring an effect statistically significant, for each set of comparisons the Bonferroni correction for multiple contrasts was used (i.e. 0.05yc, where csthe number of contrasts in the set, Kirk, 1982). All graphs of ERP waveforms depict a 420-ms epoch that includes a 120-ms prestimulus baseline and 300 ms of data following the presentation of the auditory stimulus. Each waveform has been baseline corrected by subtracting the average of the prestimulus baseline points from all data points in the ERP waveform. In all figures, the arrow indicates stimulus onset and the vertical ticks represent 100, 200 and 300 ms following presentation of the auditory stimulus. For each comparison, analyses were conducted on the sum of ERP or MMN data points across 25-ms windows within the definitional time frame of MMN. These windows were centered on 100,
125, 150, 175, 200 and 225 ms following stimulus onset. In all figures, statistically significant and non-significant effects are illustrated by black and gray bars, respectively. 3. Results The grand averages of the ERPs collected in response to standard stimuli presented in the baseline condition (blocks 1 and 2), all standard stimuli presented in the oddball condition, as well as all deviant stimuli presented in the oddball condition are presented in Fig. 1. As is customarily observed, the waveform that depicts the grand average of all ERP responses to deviant stimuli shows an enhanced negativity in the MMN window as compared to the waveform associated with the average of all standards presented in the oddball condition (oddball standards). Of relevance to the present study, the expected minimum negativity was, indeed, observed in the baseline condition presented at the beginning of the second block of trials (standards block 2). However, it
G.K. Starratt, A.J. Nash / International Journal of Psychophysiology 51 (2004) 201–208
should be noted that the waveform representing baseline standards presented at the beginning of the first block of trials (standards block 1) showed an unexpected increase in negativity in the MMN window as compared to all other stimuli. In fact, the phenomenon of interest in Fig. 1 is the relationship between the waveforms elicited by the three classes of standard stimuli. Fig. 2 depicts only the waveforms associated with uninterrupted standard stimuli presented in the baseline condition in blocks 1 and 2. The ERP response to the uninterrupted standards presented in the first block of trials (i.e. the first stimuli to which participants were exposed) shows a clear increased negativity in the MMN window as compared to uninterrupted standards presented at the beginning of the second block of trials. Analysis of the baseline corrected sum of ERP points for six consecutive 25-ms windows from 88 to 237 ms reveals significantly greater negativity for baseline standards block 1 than for block 2 from 113 to 212 ms following stimulus onset. Means, S.D., F and P values can be found in Table 1.
205
Table 1 Mean (S.D.), F and P values for the sum of data points across successive 25-ms windows for baseline standard stimuli by block Block 1 Block 2 Baseline standards Baseline standards 88–112 ms windowa y12.33 (22.21) 113–137 ms windowb y21.40 (21.32) 138–162 ms windowc y8.73 (15.84) 163–187 ms windowd y0.12 (17.91) 188–212 ms windowe 4.79 (19.81) 213–237 ms windowf 8.18 (20.86)
y3.07 y1.69 25.07 37.09 32.91 24.21
(20.13) (21.58) (18.10) (18.24) (14.99) (13.24)
Bonferroni Ps0.0008. F(1, 10)s4.62, Ps0.06. b F(1, 10)s49.84, *P-0.0001. c F(1, 10)s40.92, *Ps0.0001. d F(1, 10)s33.20, *Ps0.0002. e F(1, 10)s27.52, *Ps0.0004. f F(1, 10)s5.59, Ps0.04. a
The MMN waveforms in Fig. 3 were calculated from the waveforms shown in Fig. 1. As is commonly done, the traditional MMN waveform in this figure results from the subtraction of the
Fig. 2. Grand average waveforms recorded in response to baseline standards in blocks 1 and 2 (from Fig. 1). Arrow indicates stimulus onset. Data recorded at frontal (Fz) electrode. Intervals showing statistically significant (Bonferroni Ps0.008) and nonsignificant differences between the waveforms are indicated with black and gray bars, respectively.
206
G.K. Starratt, A.J. Nash / International Journal of Psychophysiology 51 (2004) 201–208
waveform representing the response to standards in the oddball condition (oddball standards) from the waveform related to deviant stimuli. Similarly, the alternate method MMN waveform depicts a subtraction of the ERP waveform representing responses to baseline standards block 2 from the same deviant waveform. Consequently, this figure reflects a comparison of two methods of calculating MMN. Note that, MMN is more negative in the waveform calculated using the baseline standards block 2. This difference is statistically significant (Bonferroni Ps0.01) from 163 to 237 ms. Means, S.D., F and P values can be found in Table 2.
Table 2 Mean (S.D.), F and P values for the sum of data points across successive 25-ms windows for MMN by method of calculation
4. Discussion
only for the uninterrupted string of standards presented at the beginning of the second block of trials (standards block 2). That is, even though the uninterrupted string of standard stimuli presented to the participant at the beginning of the first and second blocks of trials were identical, these stimuli, when initially presented to the participant, show a significantly enhanced negativity during
The results demonstrate that, as predicted, an uninterrupted string of standard stimuli does elicit reduced negativity in the time frame of MMN (i.e. f100–200 ms following stimulus onset) when compared to the average of all standards collected in the oddball condition. However, this was true
Traditional method Alternate method 113–137 138–162 163–187 188–212 213–237
ms ms ms ms ms
a
window y7.55 (6.48) windowb y12.75 (8.95) windowc y5.08 (12.38) 5.23 (8.93) windowd 6.28 (6.54) windowe
y12.11 y23.64 y15.20 y4.71 y4.16
(16.73) (17.38) (16.55) (11.41) (14.73)
Bonferroni Ps0.01. a F(1, 10)s1.29, Ps0.28. b F(1, 10)s8.17, Ps0.02. c F(1, 10)s11.37, *P-0.01. d F(1, 10)s9.78, *Ps0.01. e F(1, 10)s8.63, *Ps0.01.
Fig. 3. Grand average MMN waveforms calculated using traditional vs. alternate methods. Arrow indicates stimulus onset. Data recorded at frontal (Fz) electrode. Intervals showing statistically significant (Bonferroni Ps0.01) and non-significant differences between the waveforms are indicated with black and gray bars, respectively.
G.K. Starratt, A.J. Nash / International Journal of Psychophysiology 51 (2004) 201–208
this time frame (as compared to all other standard and deviant stimulus conditions), which returns to baseline levels by approximately 250 ms. Conversely, when participants were presented with the same stimuli following the first 1000 trials (and a 5-min break), the result is a greatly attenuated negativity (compared to all other standard and deviant stimulus conditions) in the same time frame. The enhanced negativity associated with the processing of initial standard stimuli in this case may reflect the additional processing resources allotted to novel stimuli by participants who were ¨ to electrophysiological methodology. Connaıve versely, the attenuated negativity associated with processing of the uninterrupted standards at the beginning of the second block of trials may reflect increased comfort and familiarity with the testing situation and the stimuli in general. That is, it is reasonable to assume that the processing of uninterrupted standard stimuli later in the experimental session would reflect some minimum level of processing associated with maintaining vigilance in a familiar environment. Consistent with these baseline differences, MMN calculated by the proposed method (deviant minus standards block 2) showed increased negativity in relation to the traditional method (deviant minus oddball standards). The significant difference between these MMN waveforms occurs between 138 and 237 ms, is in the expected direction and reflects a more theoretically conservative measure of MMN. It has been reported that the signal-to-noise ratio (SNR) in MMN can be problematic (e.g. McGee et al., 2001). The present study suggests that one way to address this problem may be to include a condition that presents an uninterrupted string of identical standard stimuli for the purposes of providing a baseline measure reflecting an absence of MMN. Use of this baseline method yields a measure of MMN that shows increased negativity in the MMN window that does not reflect simply an arbitrary inflation of this measure, but provides a conceptually rigorous index of the MMN phenomenon. It should be noted that the MMN generated by the alternate method not only reflects a larger MMN peak, but also depicts a different MMN
207
morphology. Specifically, in Fig. 2 the alternate method MMN demonstrates an extended period of MMN that lasts almost until 250 ms (i.e. almost twice as long as the traditional MMN). This later epoch observed in the proposed MMN is not seen in the traditional waveform as a result of its presence in both the deviant and oddball standard waveforms. It appears that this difference in morphology may reflect an N2 component that is present in the deviant waveform and in the waveform associated with oddball standards, but not present (or greatly attenuated) in the standards block 2 condition. Although N2 is reportedly elicited to target stimuli (e.g. Alho et al., 1994), if this does indeed represent an N2 component, it appears that there may be some vestige of this component in even unattended stimuli that is present, to an equal degree, in both standard and deviant stimuli in an oddball paradigm. As a design issue, these results indicate that the presence of negativity in the waveform is not influenced only by its immediate context (i.e. the immediately surrounding stimuli), but by its placement in the entire context of the experimental session as well. Specifically, the participants in the present experiment were exposed to two identical 1000-trial blocks, each of which began with 60 uninterrupted standard stimuli. These standard stimuli when presented in the first block, however, elicited increased negativity relative to all other standard and deviant waveforms. Of note, this negativity is increased only in the MMN window as the waveform returns to baseline with the others at approximately 250 ms. Conversely, the significantly attenuated negativity is seen only in the waveform depicting responses to uninterrupted standards when they are presented at the beginning of the second block of trials. From a design standpoint, this divergence indicates that the placement of the uninterrupted standards in the experimental situation is an important issue. In the present study, the minimum negativity was observed in response to a second set of uninterrupted standards presented midway through a 90-min testing session when participants had been informed to expect exactly the same sequence to which they had already been exposed.
208
G.K. Starratt, A.J. Nash / International Journal of Psychophysiology 51 (2004) 201–208
Under what conditions might an absolute minimum MMN be observed? First, it may be that stimulus processing will be minimized in a string of uninterrupted standard stimuli to which there has been previous exposure when the subject has been informed to expect the recurrence. However, if an unannounced, uninterrupted string of standard stimuli occurred unexpectedly in the middle of an oddball paradigm, this would likely be interpreted as a novel change in the experimental situation and would not be expected to reduce processing but, if anything, might even increase negativity associated with processing of the second set of uninterrupted standards. Second, it is possible that presenting the baseline condition even later, perhaps near the end of the session might further reduce the negativity in this waveform. However, it is also reasonable to assume that some sort of expectation effect could manifest near the end of a test session when the participant anticipates that the testing session is drawing to a close. It is likely that this type of anticipation effect would also result in increased rather than decreased negativity. While the experimental conditions for the present study may not be representative of typical MMN studies, which often utilize shorter ISIs and a smaller discrepancy between standard and deviant tones, the current procedures are consistent with conditions that have been previously reported to be appropriate for eliciting a MMN component ¨¨ ¨ (reviewed by Naatanen, 1992). Future studies may address the question of whether these results are replicable under diverse experimental conditions. In conclusion, the present results indicate that a sequence of uninterrupted auditory standard stimuli that is familiar and expected, appears to reduce the negativity associated with processing these stimuli. Consequently, standard stimuli presented under these conditions may provide an alternate baseline measure with reduced error for the calculation of MMN. The ability to improve the SNR for this component yields a theoretically conservative and authentic index of MMN, which can
increase the probability that the results of an experimental manipulation will be detectable. This may permit the design of new experimental methodologies that may ultimately yield new insights into these elusive phenomena and their relationship to fundamental theoretical questions about the nature of human cognition. References Alho, K., Woods, D.L., Algazi, A., 1994. Processing of auditory stimuli during auditory and visual attention as revealed by event-related potentials. Psychophysiology 31, 469–479. Cheour, M., Leppaenen, P.H.T., Kraus, N., 2000. Mismatch negativity (MMN) as a tool for investigating auditory discrimination and sensory memory in infants and children. Clin. Neurophysiol. 111, 4–16. Cheour-Luhtanen, M., Alho, K., Sainio, K., et al., 1996. The ontogenetically earliest discriminative response of the human brain. Psychophysiology 33, 478–481. Jasper, H.H., 1958. The ten–twenty electrode system. Electroencephalogr. Clin. Neurophysiol. 10, 371–375. Kirk, R.E., 1982. Experimental design. second ed.. Brooksy Cole, Monterey, CA. McGee, T.J., King, C., Tremblay, K., Nicole, T.G., Cunningham, J., Kraus, N., 2001. Long-term habituation of the speech-elicited mismatch negativity. Psychophysiology 38, 653–658. ¨¨ ¨ Naatanen, R., 1979. Orienting and evoked potentials. In: Kimmel, H.D., van Olst, E.H., Orlebeke, J.F. (Eds.), The Orienting Reflex in Humans. Lawrence Erlbaum Associates, Hillsdale, NJ, pp. 61–75. ¨¨ ¨ Naatanen, R., 1992. Attention and Brain Function. Lawrence Erlbaum, Hillsdale, NJ. Pekkonen, E., Rinne, T., Reinikainen, K., Kujala, T., Alho, K., ¨¨ ¨ Naatanen, R., 1996. Aging effects on auditory processing: an event-related potential study. Exp. Aging Res. 22, 171–184. ¨¨ ¨ Sams, M., Alho, K., Naatanen, R., 1983. Sequential effects on the ERP in discriminating two stimuli. Biol. Psychol. 17, 41–58. ¨¨ ¨ Sams, M., Alho, K., Naatanen, R., 1984. Short-term habituation and dishabituation of the mismatch negativity of the ERP. Psychophysiology 21, 434–441. Winkler, I., Paavilainen, P., Alho, K., Reinikainen, K., Sams, ¨¨ ¨ M., Naatanen, R., 1990. The effect of small variation of the frequent auditory stimulus on the event-related brain potential to the infrequent stimulus. Psychophysiology 27, 228–235.