INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

RESPIRATORY SENSATION IS PART of our everyday experience; for example, the sensation of dyspnea with exertion. The perception of respiratory sensation is accepted as evidence of cortical processing, but the neurological pathways involved are not precisely defined. There is a clinical need to define the neurological pathways of respiratory sensation because some patients with asthma have poor perception of the severity of their disease (19). The sensations due to added inspiratory load have been extensively studied by using psychophysical methods, and it is established that healthy humans can detect and quantify added inspiratory loads (8, 34, 35).

Somatosensory evoked potentials are small scalp signals that arise when sensory information arrives at the cerebral cortex, and they are used clinically to objectively assess sensory pathways. An event-related scalp potential that is evoked by inspiratory loading was described in 1986 (10), and this could be a useful tool in exploring the pathways of respiratory sensation. Since the initial report (10), a variety of respiratory stimuli has been used to evoke scalp potentials, and these have included added inspiratory resistance (3, 23), total airway occlusion (10, 16, 18, 29), and subatmospheric pressure pulses (24, 32). The first peak in the potential described by Davenport et al. (10) was positive; it had a mean peak latency of 60 ms and was labeled the P₁. This initial peak was attributed to the arrival of sensory information at the cerebral cortex, analogous to the first cortical component of a somatosensory evoked potential (10). The later peaks in this respiratory-related evoked potential were labeled N₁, P₂, and N₂, with peak latencies of 117.0, 169.7, and 211.6 ms, respectively. These later peaks were assumed to reflect the cortical processing of the sensory information evoked by the inspiratory load, and subsequent studies have supported that hypothesis (32).

This respiratory-related evoked potential is novel because it arises in response to a stimulus that also evokes a muscle response (4, 26). In addition, this potential has some features that are unusual for a somatosensory evoked potential. The first peak, the P₁, may be unusually large (32), and it requires relatively few stimulus presentations to increase the signal-to-noise ratio to a point at which a potential is clearly distinguishable from background noise (29). In addition, temporally related noncephalic potentials have been described (24). Some of these features may be related to the choice of reference electrode. Davenport et al. (9) compared recordings at C₃ and C₄ using both C_Z and linked earlobes as reference electrodes. When they (9) used a linked earlobe reference electrode, they found that the P₁ had a shorter peak latency and a greater amplitude (by ~1 µV), and it was slightly narrower. Thus the greater common mode rejection provided by the C_Z reference electrode had removed or attenuated a signal that was present at the earlobe. Other evidence of occlusion-related activity at sites that do not overlay the somatosensory cortex was a report of evoked potentials that were measured between the shin and the shoulder (24). Strobel and Daubenspeck (32) demonstrated an unusually large early scalp potential in one subject (amplitude of ~40 µV) when a subatmospheric pressure pulse was used as the stimulus. They (32) used an A₂ (right ear) reference electrode that would provide less common mode rejection than a C_Z electrode, and this may explain the large signal. Revelette and Davenport (29) used a C_Z reference electrode and showed that as few as 32 stimulus presentations were required to increase the signal-to-noise ratio to the point at which the evoked potential, including the P₁, could be distinguished from background noise. This contrasts with the 250 stimulus presentations that are typically required for a somatosensory evoked potential to be distinguished from background noise (22). The P₁ is thus an unusual evoked potential in terms of both the evoking stimulus and some features of the response.

The present study was undertaken to determine the relationship between the inspiratory muscle response to airway occlusion and the evoked potential recorded on the scalp. It was speculated that activity arising in the inspiratory muscles may contribute to the signal recorded on the scalp. Skeletal muscle has been described as the "most notorious" of the physiological artifacts in somatosensory evoked potential recording (28), and there are previous examples of scalp signals with a myogenic rather than a cerebral origin (1, 2, 6). The muscular response to transient inspiratory loading arises from the typical inspiratory muscles and also from oral and pharyngeal muscles. The intercostal muscles, the diaphragm and scalenus anterior, show an initial short latency inhibitory response (mean peak latency, 34 ms) followed by an excitatory response (mean peak latency, 105 ms) (4, 26, 27). Other muscles that respond to inspiratory loading include the masseter (20) and genioglossus (21). The genioglossus response has an onset latency of 34 ms (21). The latencies of these responses are similar to those of the P₁, but the relationship between the scalp potential and these myogenic responses is unknown.

In the present study, electrical activity was recorded at C₃, C₄, F₃, and F₄ referenced to C_Z, the scalene electromyogram (EMG), the electrooculogram (EOG), and the masseter EMG. Signals evoked by occlusion could be recorded from widespread sites, and there was a close, temporal relationship between the muscle signals and the scalp response. The onset latencies and the power spectra of the averaged electroencephalogram (EEG) at C₃ and C₄ and the averaged scalene EMG were similar. The onset latency of the masseter response was consistent with it being a movement-related artifact in most subjects. The EOG and the frontal EEG responses were extremely variable and were assumed to reflect facial muscle activity. It is concluded that there are widespread myogenic responses to airway occlusion and that these are a potential source of artifact in experiments of this nature.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Data were collected from six male and six female healthy volunteers. None of them were smokers, and none gave a history of respiratory disease. Nine subjects were studied twice. Respiratory and electrophysiological signals were recorded while the subject breathed through a purpose-built respiratory system. On a breath-by-breath basis, data were recorded as 350-ms sweeps in a fixed temporal relationship to the inspiratory stimulus. The stimulus consisted of a 200-ms midinspiratory occlusion, with unoccluded breaths as the experimental control. Subjects were asked to breathe normally and to maintain a constant respiratory pattern using visual feedback of their inspiratory flow. Data were recorded while 250 consecutive breaths were transiently occluded and then during 250 nonoccluded control breaths. During the experiments, which lasted 3-5 h, subjects were seated in a reclining lounge chair, with the upper body, neck, and arms supported. For comfort, rest periods were provided, and subjects were disconnected from the data-acquisition system and were able to move freely.

The experimental procedures were approved by the University of Auckland Human Subjects Ethics Committee, and all subjects gave written informed consent. There was no compensation for participating in the study. The mean height, weight, and age of the subjects were 171.5 ± 9.8 (SD) cm, 70.7 ± 16.4 kg, and 30.4 ± 5.0 yr, respectively.

Respiratory apparatus. The general arrangement of equipment is shown in Fig. 1. Subjects wore a nose clip and breathed via a mouthpiece through a purpose-built respiratory apparatus. The inspiratory limb of a T-shape Rudolph valve (2600, Hans Rudolph) was attached to noncompliant, reinforced plastic tubing (35-mm internal diameter and 1.4-m length) and an occlusive valve system that delivered precisely timed occlusions. The valve system was pneumatically driven and electronically controlled. Removal of a rubber stopper allowed unimpeded inspiration during unoccluded control breaths while the valve system continued to actuate. Valve actuation created a clearly audible click. Inspiratory flow and pressure were displayed continuously with the use of a Macintosh LCII computer (Apple Computer) and Scope V3.0 software (Mac Lab, Analog Digital Instruments). Inspiratory flow was obtained from a pneumotachograph (model 20172B, Hewlett-Packard) and a differential pressure transducer (model MP 45-1, Validyne Engineering). Mouth pressure (Pm) was recorded from the Rudolph valve with a differential pressure transducer (model MP 45-1, Validyne Engineering). The Macintosh computer monitor was placed in front of the subject, and the display was used as a source of feedback. After 5 min for stabilization, a 30-s recording of inspiratory flow and Pm was stored as a template using the Macintosh LCII computer. This stored template was reused during repeated experiments. Inspiratory flow and pressure were overlaid on this template, and subjects were asked to match the template. All subjects were able to maintain a constant respiratory pattern.

View larger version (26K): [in this window] [in a new window]   Fig. 1.   Schematic representation of experimental apparatus. Thick solid lines, bioelectric signals; dashed lines, respiratory signals; thin solid lines, trigger signals.

Recording system. In addition to the respiratory signals, the EEG, the EMG, and the EOG were recorded with gold disk surface electrodes (type E5GH, Grass Instruments). The impedance was measured with a Grass EZM 4 electrode impedance meter (Grass Instruments), and electrodes with impedance >5 k were reapplied. The scalp electrodes were placed according to the 10-20 system (17) at C₃, C₄, F₃, F₄, and C_Z, with a ground electrode placed on the forehead. C_Z was used as the reference electrode for the EEG. The scalenus anterior EMG was recorded from the right side of the neck by using two electrodes placed 4 cm apart over the scalenus anterior, with the superior electrode at the level of the cricoid cartilage. The scalenus anterior was selected as being representative of the inspiratory muscles (7). The masseter EMG was recorded from electrodes placed 4 cm apart over the right masseter with the inferior electrode 1 cm above the angle of the jaw. The EOG was recorded with one electrode placed at the outer canthus and the other placed above the eyebrow at the level of the pupil. The EOG electrode placement was chosen to allow detection of eye movement in either the vertical or the horizontal planes and detection of orbicularis oculi EMG activity. The position of the EMG and EOG electrodes was checked by using a storage oscilloscope to examine the signals generated by sharp inspiration (scalene), jaw clenching (masseter), and eye blink.

Data acquisition and storage. A Grass model 12B Neurodata acquisition system (Grass Instruments) was used to band-pass filter (1-1,000 Hz, for all channels) and amplify (100,000 times) the signals. A CED 1401 (Cambridge Electronic Design) signal processor was used as an analog-to-digital converter, with the sampling rate set at 2 kHz. Digitized data were stored in an IBM-compatible 386 computer for off-line averaging and analysis. Commercial software (Sigavg 5.1, Cambridge Electronic Design) controlled the data acquisition and provided signal averaging, three-point smoothing, and graphing functions. Data collection and valve closure were triggered when the integrated flow signal (Polygraph Integrator, model 7P10, Grass Instruments) exceeded a preset threshold. The triggering device was locally built, and the onset of the occlusion was manually adjusted to occur at midinspiration. A peritrigger data collection mode was used to provide a smooth baseline before the onset of the occlusion.

Off-line data analysis. Data were reviewed sweep by sweep to exclude eye movement and facial expression artifact and were then added to a buffer that stored an average signal. For subsequent analysis, an average was created from 200 sweeps. High-frequency components were removed by using three-point smoothing (Sigavg 5.1). Graphic data from occluded and unoccluded breaths were overlaid for comparison. Onset latency, peak latency, and peak-to-peak amplitude were measured by using a cursor within Sigavg 5.1. Latencies were measured relative to the point at which Pm began to fall after valve closure, and the same point was used in the control data. Power spectral analysis of a 100-ms segment of data from C₃-C_Z, C₄-C_Z, and the scalene EMG was undertaken by using a purpose-written LabVIEW program after the data had been averaged and three-point smoothed (Sigavg 5.1).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Data analysis was confined to the first 100 ms after Pm began to fall at the onset of airway occlusion. Control data were collected from a comparable time during the unoccluded breaths. No evoked potentials were apparent in the raw data, and all analyses were of signal averages.

The consistent finding in this study was an early positive deflection (P₁) at C₃-C_Z and C₄-C_Z and a scalene EMG response. These responses occurred during occluded breaths and were either absent or significantly reduced during the unoccluded control breaths. In two subjects, the P₁ was unilateral, and in one of these it was tiny (0.5 µV) and contained high-frequency noise, so that it was difficult to identify. The evoked response was apparent on visual inspection of the averaged EEG. Power spectral analysis confirmed what was apparent from the untransformed data. There was a 3- to 10-fold increase in low-frequency power at C₃-C_Z, C₄-C_Z, and the scalene EMG in response to the occlusion. An increase in power was clear in all 12 subjects, but the increase in power of the P₁ was least when the amplitude of the evoked response was low. In the subject in whom the P₁ was unilateral and difficult to identify, the increase in power was unilateral. The mean (±SD) onset latencies were 15.7 ± 3.1 ms at C₃-C_Z, 15.9 ± 2.1 ms at C₄-C_Z, and 17.6 ± 5.5 ms at the scalene EMG. The pattern of the averaged scalene EMG consisted of a positive wave following an initial negative wave, except in one subject in whom the pattern was inverted. In some subjects (e.g., Fig. 2), the initial 100 ms of data from C₃-C_Z and C₄-C_Z were strikingly similar to the scalene EMG. The onset of the P₁ was identified by a small negative deflection in four of the subjects, and in the others the onset was from the baseline. The onset latency of the P₁ was within 5 ms of the onset of the scalene EMG response in 11 of the subjects. The subject whose P₁ contained high-frequency noise was excluded from this analysis because the onset was difficult to define. The amplitude of the scalene EMG response was generally much greater than the amplitude of the P₁. The mean (±SD) peak amplitudes were 1.7 ± 0.6 µV at C₃-C_Z, 1.9 ± 1.0 µV at C₄-C_Z, and 6.1 ± 2.8 µV at the scalene EMG. There was considerable variability in the morphology of the scalene EMG, particularly in the duration of the initial negative deflection. The mean (±SD) duration of this negative deflection was 34.0 ± 15.01 ms. The C₃-C_Z, C₄-C_Z, and scalene EMG responses were reliably evoked, both on the same day and on different days. Data from four subjects showed an early positive peak in C₃-C_Z and C₄-C_Z during unoccluded control breaths. These potentials in the control data were generally smaller (mean 0.67 µV) and had a longer latency (mean 60.5 ms) than the P₁ from occluded breaths. Figures 2 and 3 illustrate an early peak in the control data.

View larger version (27K): [in this window] [in a new window]   Fig. 2.   Averaged data from subject 5. Thick lines, data from occluded breaths; thin lines, data from unoccluded control breaths. Vertical dotted line shows onset of occlusion. Experimental data were acquired from 250 consecutive breaths with midinspiratory occlusion and from 250 unoccluded control breaths. Data from C3-CZ, C4-CZ, and the scalene electromyogram (EMG) were similar in this subject. In addition, this subject was atypical in showing an evoked potential in the frontal electrodes. EOG, electrooculogram; P1, 1st peak.

View larger version (26K): [in this window] [in a new window]   Fig. 3.   Averaged data from subject 2. Thick lines, data from occluded breaths; thin lines, data from unoccluded control breaths. Vertical dotted line shows onset of occlusion. Frontal EEG channels show a typical response that lacks the appearance of a somatosensory evoked potential. The masseter response is similar to the mouth pressure in terms of latency and termination.

Data from F₃-C_Z and F₄-C_Z were highly variable, and there was no consistent morphology. Similar variability was noted for the EOG. Because of this variability, no further analysis was undertaken by using the frontal EEG or the EOG data. The data illustrated in Fig. 2 are unusual in that a frontal evoked potential is apparent.

The onset and offset of the masseter EMG response coincided with the occlusion, as recorded in the Pm, in nine subjects. In these subjects, the mean onset latency of the masseter signal was <3 ms, and the response was interpreted as movement artifact. Figure 3 illustrates the typical masseter response, showing onset and offset with close relationship to the Pm. In the other three subjects, the onset latencies were 16, 28, and 48 ms. In these three subjects, the masseter signal may have been the masseter myogenic response to the occlusion.

The reliability of data was assessed with repeated measurements in nine subjects. Six subjects had repeated recordings during a single experimental session, and three others were studied on two occasions separated by 2 wk. The degree of similarity was greatest when repeated experiments were conducted in one experimental session.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The present study shows that evoked responses can be recorded at multiple sites on the head and neck during the first 100 ms after a midinspiratory occlusion. This is not surprising, because such scalp signals have been previously described (9, 10, 18, 23, 29, 32, 33) and because many muscle groups in the head, neck, chest wall, and trunk are active during inspiratory loading (5, 7, 20, 21). The present study is unique in that it examined the scalp signals and myogenic responses simultaneously. It was found that the onset latency of signals recorded at C₃-C_Z, C₄-C_Z, and scalenus anterior was similar and that the scalene signal had a greater amplitude. The masseter EMG had an onset latency and general morphology that were similar to the Pm in nine subjects, and thus it is assumed that it represents movement artifact in those subjects. In particular, it was viewed that the onset latency (mean <3 ms) was too short to be a physiological response. The frontal EEG and the EOG showed great variability, and this was interpreted as facial muscle activity. It is concluded that the inspiratory muscle response to airway occlusion has a close temporal relationship to the evoked scalp signal and, therefore, may be a source of artifact in experiments of this nature. The present experiment did not quantify the extent to which the myogenic response is represented in the scalp signal, and this needs to be determined so that the early components of the scalp signal can be interpreted with confidence.

It is assumed that the scalp signal reported in the present study is the same as that reported by other workers (9, 10, 18, 23, 25, 29, 32, 33). The assumption that the P₁ is cephalic in origin was initially based on its morphological similarity to other evoked potentials (10). Subsequent studies have supported the conclusion that the P₁ is cerebral in origin, and these have included two scalp mapping studies (9, 25) and experiments searching for evidence of artifact (23, 32). Straus et al. (31) expressed concern that the scalp signal evoked by phrenic nerve stimulation may be a myogenic potential arising from the diaphragm; however, on the basis of the weight of evidence, they (31) concluded that this was improbable.

As part of a study that was mapping the scalp distribution of this occlusion-evoked scalp potential, Davenport et al. (9) compared the potentials that were recorded at C₃ and C₄ when a joined earlobe reference electrode was used with simultaneous recordings using a C_Z reference. When a C_Z reference electrode was used, the P₁ had a longer peak latency, a lower peak amplitude, and a longer duration, although the onset latency was unchanged. These differences in the P₁ were attributed to the early frontal negative peak that had a mean peak latency 13 ms after the P₁. This early frontal negative peak was apparent at C₃ and C₄ when a joined earlobe reference electrode was used but was not observed when a C_Z reference was used. The use of a C_Z reference electrode was thus providing increased common mode rejection that reduced the effect of a signal that was present at the joined earlobes.

Logie et al. (25) undertook a scalp-mapping experiment to determine the source dipole location of scalp signals that were evoked by airway occlusion. This study, like that of Davenport et al. (9), used a joined earlobe reference electrode, and the results largely confirmed the earlier work of Davenport et al., again showing an early frontal negative potential and an early parietal positive potential. Logie et al. (25) added to the earlier work by noting an even earlier positive potential (latency 15-28 ms) that had widespread scalp distribution. This widespread distribution was interpreted as indicating a noncortical origin to this early part of the evoked signal, although no likely source for this component was offered.

Knafelc and Davenport (23) addressed the possibility of myogenic artifact in the context of occlusion-related evoked potentials. They did this by recording signals between the spinous process of C₁ and C_Z, between the spinous processes of C₁ and C₇, and between the spinous processes of C₇ and T₁₂. In the recordings from the spinous processes, they found no peak that was comparable to the P₁, and thus they concluded that there is no myogenic component to the scalp signal. However, they were recording over nonmuscular bony structures, and, if there was a distant, occlusion-related myogenic signal, then it would have been recorded at all their sites simultaneously and would have been eliminated by common mode rejection. No EOG evoked potential was noted.

Earlier reports (9-11, 23, 29, 33) have assumed that a cephalic reference electrode would totally eliminate a noncephalic signal and that the scalp topography confirms a cortical origin for the signal (9). Certainly, the more closely together that scalp electrodes are placed, the greater the common mode rejection and thus the greater the confidence that the scalp signals do arise in the vicinity of the electrodes. However, there is no guarantee that this is the case. Signal averaging techniques are a powerful way of allowing coherent signals to be recognized in the presence of larger, but noncoherent signals. Thus a large noncephalic signal that is synchronous with the averaging trigger may be recorded on the scalp, even though it is only a small scalp potential. For example, Gaeta (14) showed that the electrocardiogram can be recorded at C₃-C_Z and C₄-C_Z by timing the EEG acquisition with the QRS and then averaging the EEG for 256 heartbeats.

The present study shows that airway occlusion evokes electrical activity from multiple sources, and thus the scalp signal may include both myogenic and neural signals. This does not necessarily challenge the existence of a respiratory-related evoked potential but does add the caveat that interpretation of the P₁ needs to be made with caution, because the scalp signal is likely to contain interacting components arising from multiple sources. The morphology of any scalp signal is determined by the vectorial sum of the contributing dipoles. The P₁ onset latency is similar to that of the inspiratory muscles, and these muscles respond almost simultaneously as a group during inspiratory occlusion (4). In addition, there are responses from accessory muscles, including genioglossus, masseter, and sternomastoid (20), and the chest wall and truncal muscles (5). The scaleni are representative inspiratory muscles, but they are relatively small. The bulk of the inspiratory muscle mass is provided by the intercostal muscles and the diaphragm. Because the main inspiratory muscles respond as a group, a large myogenic signal is expected in response to airway occlusion, and the onset latency of this response is similar to that of the P₁.

An explanation is required to reconcile this argument with the observation that two subjects in the present study had a unilateral response and that, in one of these subjects, the response was very small. If the myogenic response to airway occlusion is contributing the scalp signal, then the determinants of surface EMG need to be considered (12, 30). These include the distance from the signal, the orientation of the electrode pair relative to the dipole of the signal, and the magnitude of the source signal. Uncontrolled variation in one of these could be responsible for the unilateral and small response noted.

Other responses examined in the present study were the frontal EEG (F₃-C_Z, F₄-C_Z), the EOG, and the masseter EMG. An evoked response was not reliably recorded at the frontal EEG and the EOG channels. These signals contained high-voltage activity that was quite variable and, therefore, was interpreted as arising from the facial muscles. The absence of a frontal EEG contrasts with the findings in the mapping study of Davenport et al. (9). This contrast may be due to the difference in the reference electrode. Davenport et al. used a joined earlobe reference electrode, whereas a C_Z reference was used in the present experiments. The masseter response in the present study was interpreted as a movement artifact in some subjects and as the masseter response to inspiratory loading in others. Nine subjects showed an almost instantaneous masseter response, with a mean onset latency of <3 ms and a morphology that was similar to the Pm signal. Because of this short latency, this pattern of recording from the masseter was assumed to represent movement artifact. In the other three subjects, the onset latencies for the masseter responses were 16, 28, and 48 ms. These responses did not resemble the Pm trace and may have been the myogenic response from the masseter that has previously been reported (20).

Additional sources of artifact that were considered include microreflex and movement. Microreflexes are small voltage signals that are evoked from skeletal muscle in response to a sensory stimulus and that are only apparent after signal averaging. Microreflexes arise from most muscle groups, and those that arise from muscles attached to the head can be difficult to distinguish from evoked cortical responses (1). These signals increase in amplitude with increasing muscle tone and decrease with relaxation. Butler et al. (4) showed no evidence of microreflex during an experiment similar to the present study. Microreflex was, therefore, not studied in the present experiment. On the basis of previous reports (23, 32), movement artifact is unlikely to make a major contribution to the P₁, and, therefore, this was not examined in the present study.

An evoked potential was present in the control data of four subjects, including the subject who showed a very small and unilateral P₁. The cause of the potential in the control data was not studied in the present experiment. Using identical equipment, Gaeta (14) concluded that the response in the control data was an auditory evoked potential, because it was absent in one subject who had sensorineural deafness, and this subject showed a typical occlusion-related potential. An alternative explanation for the evoked potential in the control data may be that it was caused by the resistance provided by the inspiratory limb of the breathing system. Assuming laminar flow, Poisseuille's Law was used to calculated the resistance to inspiration as 0.04 cmH₂O · l¹ · s. Although this resistance is low, it may explain the evoked potential in the control data.

The present study is unique in that it simultaneously recorded the EEG and EMG responses to transient midinspiratory airway occlusion. Evoked signals were recorded from multiple sites on the head and neck. The responses that were recorded at C₃-C_Z and C₄-C_Z and the scalene EMG had similar onset latencies. It is concluded that the early inspiratory response recorded on the scalp may contain activity from the inspiratory muscles and that this needs to be considered as a source of EEG artifact in experiments of this nature.

This study emphasizes the possibility that the myogenic response to airway occlusion could contribute to the somatosensory evoked potential that is evoked by the same stimulus. The present study neither proves nor quantifies this hypothesis. Nonetheless, myogenic artifact is a well-recognized problem in evoked potential recording, and it would be difficult to think of a more likely scenario in which this problem may arise than situations where a myogenic response is an inevitable consequence of the stimulus.

This controversy needs to be resolved. A similar problem arose in the past when there was a need to distinguish between somatosensory potentials evoked by peripheral nerve stimulation and those parts of the sensory nerves' action potentials that could be recorded on the scalp (13). The potentials that arose outside the brain demonstrated onset and peak latencies that were identical at all cephalic sites when a noncephalic reference was used (such as the dorsum of the hand) and these "far fields" were "virtually eliminated" when a cephalic reference electrode, such as F_Z, was used (13). By using this approach, the present controversy could be resolved by undertaking a mapping study using a distant (noncephalic) reference electrode, such as the dorsum of the hand, and then recording occlusion-related potentials over the spinous processes, the inspiratory muscles, and the scalp. These recordings would then be compared with a simultaneous scalp map recorded with reference to C_Z. If the occlusion-related EMG signal is represented on the scalp, then it should be more prominent when a noncephalic reference electrode is used, and it should undergo significant attenuation when a C_Z reference is used. Thus EMG and cortical components of this scalp signal could be differentiated.