Standard disk electrodes, well attached on the skin and scalp surface, are suitable for most recordings; needle electrodes can also be used. When recording SEP one usually aims to study during the same run peripheral, spinal, brainstem, and early cortical SEPs. Peripheral nerve volleys of action potentials as well as conducted and segmental spinal SEPs can be recorded by placing cutaneous electrodes along the route of the fiber tracts or close to the fixed generators of segmental spinal responses (near-field recording). Electrodes placed on the scalp pick up both SEPs generated in the cortex and thalamocortical fibers, which are picked up as near-field responses located in restricted areas and far-field positivities reflecting the evoked activity generated in peripheral, spinal, and brainstem somatosensory fibers.

The SEP literature is full of discussions about the most appropriate site for the reference electrode to record each of the components of the response; in fact the knowledge of the field distributions and of supposed source locations and orientations is enough for a correct choice of the recording derivation. Considering the field distribution the optimal recording condition is in theory that in which the reference is not influenced by the activity under study. Most of the far-field potentials that have proved to be clinically useful are widely distributed over the scalp. Consequently, they reach their maximal amplitude in referential recordings in which the reference electrode is placed in a noncephalic position and they are canceled, or drastically reduced, when both recording electrodes are placed on the scalp. Furthermore, a noncephalic reference common to all channels is also adequate for all near-field recordings, and there is no theoretical argument against the recommendation of using this type of montage in all situations. However, the electrical and physiologic (electrocardiogram, electromyogram [EMG], etc.) noise level may be increased by the long distance between the active and reference electrode in noncephalic
reference montages. Conversely, differential amplification between the two maxima of a dipolar potential field on the scalp produced by an identified single generator increases the signal voltage and reduces the noise level. Therefore, montages proposed for routine recordings often stem from a compromise between (i) the theoretical optimum represented by a common noncephalic reference and (ii) practical considerations in terms of pertinent components to be studied in a given clinical condition and electrical environment.

The routine four-channel montages proposed in the IFCN guidelines (8) explore the afferent peripheral volley (in the supraclavicular and popliteal fossa for upper and lower limb, respectively), the segmental spinal responses at the neck and lumbar spine levels, as well as the subcortical far-field and early cortical SEPs using scalp electrodes placed in the parietal and frontal regions for upper limb SEPs and at the vertex for lower limb SEPs. Since knowledge of the field distributions and source locations is a prerequisite for a correct choice of derivations, recording montages will be proposed after the description of normal SEP components.

Multiple-channel recordings (24, 32, 64, or more) permit assessing the scalp topographic distribution of SEPs; they can be obtained using a cap with electrodes placed according to the modified 10- to 20-electrode system for SEPs or whole-head magnetoencephalography (MEG) devices for SEFs. These multiple-channel recordings, eventually assisted by mapping based on computerized voltage or current fields interpolation, have brought invaluable information on the spatial distribution of each component of the response; moreover, they are indispensable for calculating the evoked fields sources coordinates in a spherical or MRI-based realistic model of the head.

The position of the ground electrode is more crucial for SEPs than for the recording of any other type of EPs, due to the electrical artifact produced by the stimulation. To minimize the electrical artifact produced by the stimulation, the ground electrode should be placed on the stimulated limb between the stimulation site and the recording electrodes. Flexible metal strips covered with saline-soaked cloth wrapped around the limb close to the stimulus site are recommended. Electrically isolated stimulators allow the use of a ground electrode on the head, which is adequate for eliminating artifact related to the electrical main and radiofrequency interference.

Electrical stimulation of large-diameter myelinated fibers of peripheral nerves produces a dorsal horn (DH) potential with a posterior-negative and anterior positive dipolar field perpendicular to the cord axis. This field distribution of DH potentials has been demonstrated by direct spinal recordings at all segmental levels of the cord (Fig. 48.2), and a comprehensive atlas of these responses is now available (49). The cervical N13 potential is
recorded at the posterior neck, with a maximum voltage at the level of C5 to C7 spinous processes and decreases in amplitude at more rostral or caudal electrode positions. N13 does not show any latency shift between C6 and C2 recordings in normal subjects (50, 51 and 52). When recorded anterior to the cord by an electrode placed at the anterior aspect of the neck (50,53), the cervical DH response is recorded as a spinal P13 positivity. This polarity reversal has been demonstrated with esophageal (54) and epidural (55) recordings and also by direct recordings from the surface of the cervical cord (56,57). Both N13 and P13 spinal components are reduced in amplitude to the same degree when the stimulation rate is increased over 10/sec (48), both persist in brain death or in cervicomedullary lesions (53,58,59) and both are selectively reduced or abolished in lesions of the cervical cord gray matter sparing the spinal somatosensory tracts (60). Thus the most likely generator of the N13-P13 cervical potentials is the compound segmental postsynaptic potential triggered in the DH gray matter by the afferent volley in fast conducting myelinated fibers. Direct recordings on the cord surface have shown that the characteristics of N13 in humans are very similar to those of the N1 spinal potential described in animal studies, which reflects the postsynaptic neuronal response of DH laminae IV and V neurons to inputs conveyed by group I and II peripheral afferent fibers (61, 62 and 63).

In noncephalic reference recordings the N13 often has a small voltage. A cervical transverse derivation between two electrodes located, respectively, over the C6 spinal process and above the laryngeal cartilage is the most adapted for recording the spinal N13-P13 segmental response (Fig. 48.3). Substraction of the
anterior cervical (AC) positivity from the posterior neck negativity increases the signal-to-noise ratio and has the advantage over the more conventional C6-Fz montage of recording a segmental DH response that is not contaminated by SEP components generated above the foramen magnum (see below).

Apart from its field distribution, a second characteristic of the segmental N13 potential is that it shows amplitude reduction, rather than latency prolongation, when the dorsal afferent volley is time dispersed, or when the number of responding DH neurons is reduced. For diagnostic purposes the limitation of N13 amplitude measurements is that distribution of this parameter in a group of normal subjects does not show a gaussian profile. Therefore, the lower limit of normal N13 amplitude values estimated by the mean minus 2 to 3 standard deviations (SDs) is statistically meaningless. The best way to evaluate the N13 amplitude is to calculate the ratio between the N13 amplitude and that of the dorsal root N9-P9 deflection that immediately precedes N13 in transverse cervical montage recordings (60). The distribution of the log10 (N13-N9) amplitude ratio has a gaussian profile in normals. The normal lower limit of the normal N13-N9 ratio, defined as the mean value minus 2.5 SD, is of about 1.2. The N9-P9 deflection shows good test-retest reliability and its amplitude closely reflects the incoming volley in the cervical roots, which is likely to be unaffected in spinal cord diseases. Moreover, the N13-P9 amplitude ratio remains stable when tested at stimulus intensities between motor threshold and twice the motor threshold (60). Runs of 1000 stimuli delivered at 2/sec are appropriate in most cases to obtain well-defined cervical potentials using either a noncephalic or AC reference site.

It has been known for many years that a negative potential peaking at a latency of about 13 msec can be recorded intraoperatively at the dorsal aspect of the cervicomedullary junction (64, 65, 66, 67 and 68). On the skin surface, however, it is not easy after median nerve stimulation to separate this potential at upper cervical level from the segmental cervical N13, so that the former can be interpreted as a rostral spread of the latter. In particular, there is no evidence of a latency increase of the median nerve N13 negativity from the Cv6 to Cv2 level in noncephalic reference recordings, although in bipolar recordings along the neck the mean N13 peak latency was found to be longer at rostral than at caudal levels (69). An elegant insight to this latency problem was provided by Zanette et al. (70), who showed that the N13 recorded at upper cervical level peaks 0.8 msec later than the lower cervical N13 after ulnar nerve stimulation, but not after median nerve stimulation, in noncephalic reference recordings. The greater distance from the dorsal root entry zone to the cuneate nucleus for ulnar nerve than for median nerve afferent fibers was proposed to explain this latency shift. Araki et al. (71) have addressed the question of separate generators for the lower and upper cervical N13 potentials by using a conditioning-test paired stimulus paradigm for the stimulation of the median nerve. They confirmed that at ISIs of 4 to 18 msec the test of the lower cervical segmental N13 to the second stimulus was attenuated by 2% to 34% when compared to the control response, as previously reported by Iragui (72) and by El Kharoussi et al. (22) after stimulation of median nerve and the second and third fingers, respectively. More unexpected was the finding that, while the scalp P13-P14 showed the same attenuation as the lower cervical N13, the N13 recorded at the CV2 level showed an increase of 4% to 25% at these ISI. The upper cervical N13 is supposed to be generated by the presynaptic volley in the DC fibers close to the cuneate nucleus, but the difficulty of separating it from the DH N13 lessens its clinical reliability in routine clinical recordings (see Ref. 73, for a review).

The nuchal N11 potential is recorded all along the posterior aspect of the neck, where it usually appears to encroach upon the ascending slope of N13. In noncephalic reference recordings its onset latency increases from C6 to C1 spinal processes by 0.9 ± 0.15 msec (50,52). This bottom-up shift of N11 onset latency suggests that N11 is generated by the ascending volley of action potentials in the DCs of the cervical spinal cord. As first reported by Cracco (74), it is often difficult to differentiate the N11 from the following N13 component. This seriously hampers the use of N11 in clinical practice, which is limited to cervical cord lesions that selectively obliterate the N13 potential. In direct recordings on the cervical cord surface during surgery, N11 appears as a fast polyphasic component that overlaps in time with the slower N13 segmental postsynaptic potential.

On scalp noncephalic reference recordings with a highfrequency filter over 1000 Hz and a fast sampling rate (bin width < 200 µsec), three or four stationary positivities are consistently observed with a wide distribution and a mediofrontal predominance. In normal adults these potentials peak with mean latencies of 9, 11, 13, and 14 msec and are labeled P9, P11, P13, and P14, respectively, according to the polarity-latency nomenclature. There is some interindividual variation in the waveform of the two latest P13 and P14 potentials since either the P13 or P14 peaks can predominate. The amplitude of these far-field positive potentials are not affected when the stimulation rate is increased up to 50 Hz (48).

The P9 potential is picked up at the neck as well as on the scalp. It reflects the afferent volley in the trunks of the brachial plexus in axilla and supraclavicular fossa (75,76). Its peaking latency varies with arm length and its onset latency, but not its peaking latency, is influenced by the respective positions of arm and trunk (77).

The P11 potential reflects the ascending volley in the fibers of DCs at the cervical level. Desmedt and Cheron (50) first reported that, after median nerve stimulation at the wrist, the P11 potential begins in synchrony with the cervical N11 potential at the C6 level, which is close to the dorsal root entry zone in the cervical cord. The P11 potential is not recorded in about 20% of normal controls, and thus the clinical significance of its absence in patients is questionable. Conversely, its persistence when later scalp components are abnormal is a reliable indicator of preserved DC function in patients with lesions located in the medulla oblongata or at the cervicomedullary junction.

The P13-P14 far-field potentials are consistently recorded in normal subjects (see Ref. 78, for a review). Both may be of similar amplitude, or either of them may be the larger of the two (Fig. 48.1). In some subjects P14 is hardly visible as a notch on the ascending phase of P13; in others P13 and P14 cannot be differentiated. In the same individual the morphology of the P13-P14 complex can display some degree of side-to-side difference, which creates difficulties for interpreting left-right asymmetries in patients. P14, but not P13, always peaks later than the cervical segmental N13 potential (79). Since the P13-P14 complex is picked up at the earlobe with a lesser amplitude than in the frontal region of the scalp, it can be recorded with a scalp-earlobe montage (80). The P13-P14 complex is the only scalp farfield SEP that is reduced by interfering stimulations, such as vibrations, applied to the hand on the stimulated side (48). This suggests that it is generated after the synaptic relay in the nucleus cuneatus (NC). Patients with thalamic lesions usually show a normal P14, suggesting that this component originates at subthalamic level (80, 81 and 82). Conversely, in cervicomedullary lesions (53,59) or in brain-dead patients (58) the P14 potential is absent. These findings suggest that the generator of P14 is situated above the level of the foramen magnum.

Simultaneous scalp and nasopharyngeal SEP recording in deeply comatose and brain-dead patients brought some information concerning the origins of the P14 potential (83). In patients with signs of upper brainstem dysfunction this author observed that the P14 recorded between scalp and nasopharynx is lost while an earlier P13 positivity persists in scalp to shoulder traces. This observation suggests that the P14 recorded between the scalp and the nasopharynx reflects the activity of medial lemniscus pathways at the upper part of the brainstem, whereas the P13 might be generated at the cervicomedullary junction. A normal P13 coexisting with an abnormal P14 has been observed in patients with pontine or mesencephalic lesions (69,84, 85 and 86), supporting the hypothesis that P13 and P14 potentials could be generated at different anatomical levels in the cervicomedullary and brainstem somatosensory pathways. This view is reinforced by the observation, in noncephalic reference recordings, that the P13 positivity recorded at the nasopharynx can be preserved while the scalp P14 is abnormal in lesions of the lower brainstem (87).

The scalp and nasopharynx P13 potential is different from the segmental P13 positivity recorded at the anterior aspect of the neck. Despite their similar latencies and polarities, these two components have different distributions and origins and the segmental spinal P13 is preserved in all patients with upper cervical or cervicomedullary lesions.

Thus, according to our present knowledge the most probable mechanism generating the P13-P14 potentials are (i) the ascending volley in upper cervical DC fibers at the cervicomedullary junction or in medial lemniscus fibers in the brainstem, (ii) the postsynaptic response of NC neurons, and (iii) junctional stationary potentials related with the moving action potential volley across the border between neck and posterior fossa (88,89).

For the sake of clarity, the P13-P14 complex is often labeled as P14, a convention that is adopted here. It can be useful to calculate the P9/P14 amplitude ratio in patients. As with the I/V amplitude ratio of brainstem auditory-evoked potential (BAEPs) one compares, by measuring the P9/P14 ratio, the amplitude of a peripheral potential (P9) with that of a brainstem potential (P14). In normals the P9 potential is smaller than the P14 potential when recorded with a broad bandpass of 1.6 to 3200 Hz, and the distribution of the P9/P14 ratio has a gaussian profile (90).

This potential identified by Desmedt and Cheron (91) is a long-lasting scalp-negative shift, which immediately follows P14. In normals N18 can only be identified in the posterior parietal region ipsilateral to the stimulation, where there is no or minimal interference with cortical potentials. The N18 is a long-lasting component (about 20 msec when recorded with a 1- to 3000-Hz bandpass), and when low frequencies are cut off by filtering it can be falsely interpreted as an ipsilateral cortical N20. It is still uncertain whether the N18 negativity contains several subcomponents; several wavelets are superimposed on its plateau at fixed latencies on repeated recordings in the same individual, which favor this interpretation. When all cortical responses are eliminated by deafferentation due to a lesion of the ventroposterolateral (VPL) thalamic nucleus (92) or by direct lesion of the centroparietal cortex (79), N18 can be recorded on the whole surface of the scalp (Fig. 48.4). After
hemispherectomy, which entails retrograde degeneration of the thalamocortical neurons, the N18 potential is most often preserved (93); however, N18 can be missing, as well as the P14 potential, long after hemispherectomy including the thalamus (94) in relation with retrograde degeneration of the cuneatethalamic projections. These observations rule out the possibility that N18 could reflect a cortical or a thalamic response to lemniscal or extralemniscal inputs.

The earliest hypothesis, concerning the generation of N18, was that of a postsynaptic response of the brainstem nuclei connected with the DC nuclei to ascending inputs conveyed by ancillary fibers of the medial lemniscus. This hypothesis has been supported by intracerebral SEP recordings in humans showing a stationary negativity between the upper pons and the midbrain, peaking at the same latency as the scalp N18 (95). However, nasopharyngeal recordings (96) rather suggested an N18 origin in the medulla oblongata and more precisely in the NC. Potentials recorded dorsal and ventral to NC in humans (68) show a biphasic waveform that is very similar to that of NC responses to peripheral nerve stimulation in cats. The first phase of this response in cats represents the postsynaptic potential of NC relay neurons to the DC afferent volley, while the second one reflects the presynaptic inhibition of DC fibers terminals in NC (65). This inhibition consists in a feedback depolarization of DC terminals, triggered by the afferent volley in DC fibers collaterals and mediated by NC interneurons. To date it is accepted that N18 reflects a similar phenomenon in the human brainstem (see Ref. 97, for a review). Several arguments lend substance to this analogy: (i) The distribution and polarity are similar in cat and human recordings. (ii) The presynaptic inhibition of DC terminals in cats, like the human N18, lasts for several tens of milliseconds, probably because it is mediated via a polysynaptic chain of interneurons. (iii) The N18 voltage is unaffected by interfering vibrations applied on the hand on the stimulated side (98), contrary to the P14 potential, which is attenuated (48); this finding fits with the inference that N18 reflects inhibitory activity upon DC terminals.

In normal subjects noncephalic reference recordings of scalp SEPs show two positivities of cortical origin that are superimposed upon N18 in the central and frontal regions contralateral to stimulation, namely central P22 and frontal P20 (see below). Thus the early portion of N18 appears as a negativity preceding the onsets of P20 and P22 in the midfrontal and central regions, respectively. These frontal and central negativities have been considered as genuine cortical responses and labeled as N15, N16, N17, or N19 components (99,100). However, superimposition of traces recorded in the posterior parietal region ipsilateral to stimulation, where N18 is not contaminated by contralateral cortical components, clearly shows that there is no actual scalp negativity other than N18 and N20 within the first 22 msec following stimulation. The N18 plateau recorded in the parietal region ipsilateral to stimulus can serve as a baseline to assess the voltage and the shape of cortical components peaking in the 20- to 40-msec latency range. Since most of the N18 is picked at earlobe, the use of the earlobe ipsilateral to stimulation bypasses this baseline problem by eliminating the N18 negative shift in the resulting waveforms.

These cortical SEPs are peaking in the 18- to 35-msec latency range and are all obtained with optimal voltage in response to stimuli delivered at a slow rate (less than 2/sec). They are recorded on the scalp in the parietal region contralateral to stimulation and in a large frontocentral area, mostly contralateral to stimulation. They are superimposed on the widespread N18 in noncephalic reference recordings. Although the scalp distribution of these potentials shows substantial variations between subjects, their profiles and voltages are relatively symmetrical in a given individual, so that side-to-side differences in a group of normal subjects can be used as norms for assessing interhemispheric asymmetries in patients. For latency measurements, calculating the intervals from the peak of the far-field P14 to the peak of each of the cortical components permits eliminating interindividual variability related to arm length and body height differences.

Though studies in normals and patients converge on the conclusion that responses recorded in the parietal region originate in the somatosensory area SI, there is still some controversy as to whether some of the early cortical SEPs recorded in the frontal region might be generated in the prerolandic cortex. Several clinical studies reported that frontal and parietal SEP waveforms can be selectively affected in patients with focal hemispheric lesions (19,101, 102 and 103) and suggested prerolandic generators for some of the SEPs recorded in the frontal region. These observations, based on clinical examination, abnormal SEP waveforms analysis, and computed tomography (CT) imaging of lesions were at the origin of the hypothetical sensorimotor model of early SEPs sources. This model had the advantage of fitting with the concept of a short-latency control of motor cortex by peripheral somatosensory inputs via transcortical long-loop reflexes. Conversely, direct cortical recordings in humans (see Ref. 104, for a review) supported the view that all sources of early cortical SEPs are located in the parietal cortex and more precisely in the primary somatosensory area (SI). This controversy, the terms of which are discussed in more detail later in this chapter, should not discourage clinicians from using cortical SEPs as indexes of brain dysfunction in the assessment of patients with sensory or motor disorders for two main reasons. First, correlations between a clinical deficit and an SEP abnormality are clinically pertinent facts, even if they can lead to pathogenetic hypothesis, which are not validated by other approaches. Second, in lesion studies, the extent of the lesion, as well as its functional effects, which were difficult to assess by correlating SEP waveforms with CT images of the early 1980s, are now more accessible by fusing anatomical and functional data from MRI, emission tomography, and source modeling of SEPs.

Two sets of early cortical potentials are consistently recorded in normal subjects on the scalp contralateral to stimulation. The first is made of a parietal N20 and frontal P20 dipolar field; the second is composed of a central P22 positivity (Fig. 48.5). These components are also present after finger stimulation with a

latency delay of 2 to 3 msec, as compared with median nerve SEPs, because of finger-to-wrist conduction time.

The parietal N20 and frontal P20 reflect the activity of a dipolar generator in Brodmann’s area 3b situated in the posterior bank of the rolandic fissure (105, 106 and 107) and represents the earliest cortical potential elicited by median nerve stimulation. This dipolar field distribution is also observed in magnetic recordings. All source modeling studies of electrical and magnetic fields, as well as direct cortical recordings, converge on the conclusion that the N20-P20 field is produced by a dipolar source tangent to the scalp surface reflecting the response of area 3b neurons to cutaneous inputs. This was confirmed by the finding that the N20-P20 response is not evoked by joint inputs produced by passive finger movements (20). Therefore, the polarity reversal line of the N20-P20 dipolar field recorded on the scalp or directly on the cortex and the equivalent dipolar source modeled from electric or magnetic recordings are now widely used to localize the central fissure in the assessment of patients with a lesion or an epileptogenic area located in the rolandic region.

On scalp (Fig. 48.6) and some direct cortical recordings a P22 recorded in the central region was found to peak 1 to 2 msec later than the N20-P20 potentials (91,108,109). This has been taken as an argument to consider P22 as a genuine component generated by a source distinct from that of N20-P20 (91), while others denied the existence an independent P22 (110).

Sequential spatial maps of scalp-recorded SEPs have shown that the N20-P20 dipolar field is followed in the central region contralateral to stimulation by a positive P22 field peaking 1.0 to 2.5 msec with little or no negative counterpart on the scalp (111). This spatial distribution of this P22 central positivity suggests that its source is radial to the scalp surface. Most studies of SEFs failed to confirm such a source in the central region probably because magnetic recordings are blind to magnetic fields produced by dipolar sources radial to scalp surface (112, 113 and 114).

The question of whether the radial source of the central P22 is located behind the central sulcus, in the primary somatosensory area, or in front of it, in the primary motor area, is not easy to address by scalp recordings in normals, even when assisted by dipole modeling techniques based on a spherical head model (115,116). One argument in favor of a precentral origin of the P22 potential is that it may be selectively lost in patients with hemispheric lesions who have
normal sensations but signs of upper motor neuron dysfunction and selectively preserved in patients with hemianesthesia or astereognosis and no motor deficit (19,101, 102 and 103). Some direct cortical recordings in monkeys and humans, however, failed to identify a P22 in the precentral cortex (117,118) but individualized a surface-positive potential, labeled as P25, with a source radial to the scalp surface and located at the posterior edge of the rolandic fissure in Brodmann’s area 1. In an SEP and positron emission tomography (PET) study Ibañez et al. (119) reported that at stimulus rates between 4 and 16/sec SI was the only area showing a cerebral blood flow (CBF) increase, while at stimulus rates of 0.2 and 1 Hz no CBF change was induced by electrical stimulation of the median nerve. Lastly, a dipole modeling study of SEPs using realistic head models and constraining the solution in the individual brain MRI volume supported the data from direct cerebral recordings by showing that the P22 source is located at the crown of the postcentral gyrus (120). These data from cortical recordings and modeling of scalp responses are challenged by some experimental data in monkeys: (i) short-latency unit responses to somatosensory inputs can be recorded in the motor cortex (121,122) and (ii) early median nerve SEPs are recorded in the motor cortex (123). According to the data of Nicholson-Peterson et al. (123), the earliest cortical responses to upper limb stimulation are likely to result from the approximately coincident activation of at least three dipolar sources in areas 3b, 4, and 1 with orientations both opposing and orthogonal to each other. Sources in area 3b and 4 have opposite orientations tangent to the scalp surface, while the source located in area 1, at the crown of the postcentral gyrus, is oriented radial to the scalp surface. Thus, even if it would be hazardous to imagine a strict homology between SEP components in monkey and humans, due to the considerable anatomical differences between the two species, these experimental data support the contribution of the motor cortex to the N20-P20 and P22 scalp potentials.

These potentials are recorded in the parietal region contralateral to stimulation; their peaking latencies show large interindividual variations between 24 and 27 msec. In some subjects two distinct P24 and P27 can be identified while only one of the two peaks is observed in others. This explains why, according to the polarity-latency nomenclature, the first parietal positive potential following N20 has received various labels in literature (P24, P25, or P27). These variations reflect the fact that the activities of several parietal sources overlap in time in this latency range. The P27 potential was found to be abnormal in patients with focal lesions of the parietal cortex, presenting with astereognosis and normal N30 frontal responses (19,82). A P27 source in the primary somatosensory area (Brodmann’s area 1), radial to scalp surface, would accord with these findings. The P24 potential on its own is associated with a frontal N24 potential. The dipolar source of the P24-N24 field is tangent to scalp surface and thus presumed to be oriented perpendicular to the rolandic fissure. There is no report of a lesion affecting the P24 potential with preserved N20 and P27 potentials.

The frontal potential, labeled as N30 in most clinical studies, has warranted much attention in the recent SEP literature because it shows some abnormality in patients with various types of motor disorders. It is picked up in the frontal region contralateral to stimulation; however, it often spreads to the midfrontal region and to the frontal region ipsilateral to stimulus. Its waveform shows two distinct components of which the earlier one, peaking at about 24 msec (N24), appears as notch on the ascending slope of the later one, which peaks at 30 msec (N30), and has the larger voltage of the two in normal young adults (Fig. 48.7).

There is now some consensus on the conclusions that the frontal negativity with its two N24 and N30 peaks cannot be generated by a single source (43,124,125) and that the early part (N24) of the frontal negativity corresponds to the polarity reversal of the parietal P24 potential across the central sulcus. The latter statement is supported by several observations issued from intracranial and scalp recordings:

Conversely the origin of the main N30 negativity of the frontal N24-N30 is still a debated issue. There has been a long controversy as to whether frontal N30 and parietal P27 components could have separate sources. Three early observations have supported the hypothesis that, contrary to the N24-P24 dipolar field, N30 does not represent the frontal counterpart of the parietal P27:

Based on these findings it has been proposed that the N30 and P27 potentials originate in the precentral cortex and in the SI area (Brodmann’s area 1), respectively, each of these two dipolar sources being oriented perpendicular to the scalp surface. However, the existence of any N30 potential generated in the precentral cortex has been questioned by Allison et al. (126) and McCarthy et al. (118) on the basis of direct cortical recordings in monkeys and humans. The most convincing evidence that the N30 potential may have a frontal origin stems from several observations (see Ref. 25, for a review) suggesting that the N30 potential could be related with motor programming and generated in the premotor cortex of area 6, and more precisely in the supplementary motor area (SMA):

If the origin of N30 in the SMA has not been verified, the possibility that it might be generated in the primary motor area has not been ruled out by direct recordings, or source modeling, of SEPs or SEFs. Conversely, source modeling studies supported this hypothesis (144). The movementrelated gating of N30, therefore, could reflect a similar change in motor cortex excitability both during movement and mental motor imagery. Studies using transcranial magnetic stimulation (TMS) converge to the conclusion that the excitability of the motor cortex is increased during actual movement and MMS (145,146), so that N30 amplitude would reflect the level of motor cortex inhibition. However, the fact that N30 and motor cortex inhibition are decreased during movement and motor imagery does not imply that the underlying mechanisms are the same in the two conditions. A TMS study by Ridding and Rothwell (147) suggests that, in spite of a similar effect of movement and motor imagery on N30 amplitude, activity in the circuits responsible for GABAergic inhibition of the motor cortex is reduced during voluntary contraction while motor imagery, which activates similar brain regions as overt movement but does not result in afferent input, does not produce significant changes in intracortical inhibition.

Both P45 and N60 potentials are recorded contralaterally to the stimulation in the central and the frontal regions. It is still debated whether they reflect the activation of associative cortical areas, via corticocortical connections, or cortical responses to inputs transmitted directly from the thalamus. Early SEP studies in patients with focal lesions have shown that P45 is preserved in patients with isolated loss of pain and temperature sensations (37,150,151). Moreover, it may persist in patients with lesion of the parietal cortex causing pure astereognosis and abnormal N20 or P27 over the damaged hemisphere (19,82). This latter finding suggests that the P45 potential might be generated by parallel ascending inputs rather than through corticocortical fibers originating from the SI generators of the N20-P27 potentials. Little is known about the N60 potential. The observation that its voltage reaches its plateau only when the intensity of the stimulus is increased up to seven times the sensory threshold has suggested that it might be generated by spinothalamic afferents, but lesion studies have hitherto failed
to confirm this hypothesis (Bastuji, Garcia-Larrea, and Mauguière, unpublished data). Recent intracranial recordings coupled with source modeling of scalp responses suggest that three distinct sources contribute to the N60: one is located in the frontocentral cortex contralateral to stimulation, and the two others are located in the suprasylvian cortex (somatosensory area SII) on both sides (152). In scalp recordings two distinct components can be identified on the basis of their latencies and scalp distributions, namely, a frontocentral N60 contralateral to stimulation and a bilateral N70 in the temporal regions. Thus, none of the scalp-negative potentials identified as N60 or N70 in the literature has been shown to be generated in the posterior parietal cortex.

The most precise available descriptions of the human SII area have long been those by Penfield and Rasmussen (153) and Woolsey et al. (154) based, respectively, on electrical stimulation and intracortical recordings of SEPs. They showed a single complete representation of the body located lateral to the end of the rolandic fissure on the dorsolateral surface of the hemisphere and extending to the cortex of the upper bank of the sylvian fissure, contrasting with the multiple body maps identified in the lateral sulcus of monkey’s brain (155,156). The clinical deficit resulting from SII lesions is also more controversial than that observed after SI lesions. While Penfield and Roberts (157) concluded that removal of SII is followed by “no obvious sensory or motor defect,” it has been reported that a lesion of the suprasylvian somatosensory cortex produces a deficit in tactile object recognition by the opposite hand (158). However, this observation, based on clinical and MRI data, has not been correlated with abnormalities of perisylvian SEPs; moreover, it has not been checked either that the SI responses were normal in these patients.

The existence of short-latency SEPs generated in SII peaking earlier than 40 msec after median nerve stimulus remains a debated issue in the literature. Lüders et al. (159), using an array of 96 subdural electrodes in a single epileptic patient, recorded responses to contralateral median nerve and finger stimulation in the inferior frontal gyrus, peaking only 2.4 msec later than SI-evoked responses that they considered as originating in the SII area. However, these authors failed to record such early responses ipsilateral to stimulation, and the measurement area was clearly anterior to the cortical zone producing longer latency perisylvian responses in human corticograms (117). Of the numerous MEG studies of SII responses to median nerve stimulation, only that of Karhu and Tesche (160) found neuromagnetic activity in the SII area starting 20 to 30 msec after contralateral stimulus, simultaneously, or even before the first activation in SI area. Lastly, recent stereotactic recordings in the suprasylvian cortex have confirmed early bilateral responses peaking at about 30 msec in area SII (161).

If the existence of early SEPs in SII warrants confirmation, that of SII responses peaking in the 60- to 120-msec latency range has been confirmed by several investigators using various electrophysiologic approaches. Because it picks up selectively magnetic fields produced by current sources located in the fissural cortex and tangent to the scalp surface (see Ref. 162, for a review), MEG was the earlier noninvasive technique that permitted individualizing the somatosensory response from the SII area, which is buried in the sylvian fissure. More than 20 years ago median nerve magnetic SEFs peaking in the 100-msec latency range were recorded with a scalp topography compatible with sources located in the parietal operculum of both hemispheres, more precisely in the superior surface of the sylvian fissure posterior to the crossing point of the rolandic and sylvian fissures (42,45). These SEFs originating from parietal opercular sources were shown (a) to be more variable from one subject to another than SI SEFs, (b) to be very sensitive to the ISI, and (c) to have a somatotopic organization (163,164). Early corticographic recordings on the surface of suprasylvian cortex also showed responses, interpreted as originating in SII, peaking in the 100- to 120-msec latency range, ipsi- and contralateral to the stimulus and sensitive to the ISI and task condition (see Ref. 165, for a review).

These latter potentials identified as SII responses can be recorded in scalp and deep recordings of median nerve SEPs. In scalp recordings P100 and N120 potentials are consistently recorded in the central and temporal regions opposite to finger stimulation, respectively (166). The field distribution of the N120 response is compatible with a generator located in the SII area contralateral to stimulus. At a latency of about 140 msec, a scalp negativity distributed over both central regions and the vertex is of maximal amplitude when the subject has his attention drawn to the stimulus and is instructed to mentally count the stimuli. This response is modulated by attention and shows the same behavior as the endogenous processing negativity originally described in response to auditory tones (see Ref. 167, for a review). Bilateral activation of SII areas could contribute to the generation of this N140 negativity, but, in the absence of confirmative source modeling or lesions studies, this hypothesis remains conjectural.

Direct intracortical recordings using electrodes stereotactically implanted in the upper bank of the sylvian fissure have shown an N60-P90 response to median nerve stimulation (Fig. 48.9) that represents one of the most robust cortical somatosensory response recorded outside the central parietofrontal cortex in humans (168,169). Contrary to that of SI responses, the amplitude of the SII potentials does not saturate at stimulus intensities of three to four times the sensory threshold but continues to increase up to and above the pain threshold (170). Ipsilateral SII responses are consistently recorded by depth electrodes; they peak 10 to 12 msec later than contralateral ones, a delay compatible with a transcallosal interhemispheric transmission. The difference between latencies of surface SII responses recorded by corticography and depth responses obtained by stereotactic recordings is unclear. It is probable that the first N60 potential of depth recording is not picked up on the cortical surface because its amplitude is far less than that of the following P90 response.

Using a whole-head 122-superconducting quantum interference device (SQUID) magnetometer, SEF sources has been modeled in the posterior parietal (46) and mesial cortex (47)
contralateral to stimulation and in both frontal lobes (171,172), which are all active in the 70- to 160-msec latency range, in response to stimuli delivered at long ISIs, in particular when mentally counted by the subject. Intracortical recordings lend some substance to these SEF dipole modeling findings (163). Scalp recordings of SEPs, however, do not allow one to identify components reflecting specifically the activities of these sources, as identified from SEF modeling, probably because the field distribution reflects simultaneous activation of multiple and largely distributed sources at these latencies. Moreover, there is no available study demonstrating abnormal SEP or SEF profiles associated with focal lesions affecting the cortical regions proposed as generators of these long-latency responses.

A CAP corresponding to the activation of tibial nerve fibers is recorded at the posterior aspect of the knee using a bipolar montage. The negative peak of this near-field potential peaks at a latency of about 7 msec in adults. This N7 potential reflects a mixed response of motor and sensory fibers, which is clinically useful to assess the function of the peripheral segment of the pathway. The afferent volley in cauda equina roots can be recorded using skin electrodes placed at the L5-S1 level and a distant reference site, for instance at the knee opposite to the stimulation.

An electrode situated on the spinal process of T12 or L1 vertebrae and connected to a distal reference electrode records a spinal negative potential peaking at 21 to 24 msec in normal subjects. According to different authors, this segmental response has been labeled N20 (173), N21 (174), N22 (175,176), N23 (177), or N24 (178). These differences in nomenclature are mostly due to differences in the mean body height of the subjects sampled for
normative studies. In what follows this response is labeled N22. After tibial nerve stimulation the lumbar N22 originates mostly from the spinal segment receiving fibers from the S1 root (173, 174 and 175,178, 179, 180, 181 and 182).

As all segmental DH responses (see Ref. 49, for a review) the N22 demonstrates a polarity reversal when recorded anterior to the cord, a field distribution consistent with a horizontal dipolar source reflecting the postsynaptic response of DH neurons to incoming inputs. The N22 amplitude is maximal close to the entry zone of the S1 root in cord surface recordings and decreases steeply without any latency shift at more rostral or caudal electrode sites (57). The N22 potential is followed by a slow positivity, known as the P wave, which reverses into a negativity when recorded at the anterior aspect of the cord (56,183) and may reflect the processes of presynaptic inhibition affecting the primary afferent fibers in the DH. In intraoperative direct recording of the cord, several fast negativities are superimposed on the ascending slope of the N22; they reflect the action potentials ascending in the DCs. In most skin surface recordings these negativities cannot be individualized.

When the reference electrode is not situated on the axis of propagation of the peripheral ascending volley, the N22 potential is preceded by a small positivity peaking around 17 msec. This P17 positivity is a far-field potential originating in lumbosacral plexus trunks, which can also be recorded occasionally on the scalp when the reference electrode is placed at the knee (or iliac crest) on the nonstimulated side (177,178).

With appropriate noncephalic reference montages widespread far-field positivities, other than the P17 described earlier, can be recorded on the scalp that peak before the onset of the earliest cortical potential in the 25- to 32-msec latency range (177,178,184). Only the latest of these positivities, identified by Yamada et al. (177) as the P31 potential, is consistently recorded in normals after tibial nerve stimulation at the ankle. This observation was confirmed by several investigators who labeled this potential as P28 (185,186), P30 (187, 188, 189 and 190), or P31 (178,191). The latency of this potential varies according to body height and is labeled P30 in this chapter according to the author’s normative data.

The utility of the P30 potential has been validated for clinical applications of tibial SEPs in patients with spinal cord and brainstem lesions (192,193). The P30 potential is widely distributed on the scalp with a predominance in the frontal region (187,188) and therefore is drastically reduced in scalp-reference recordings (191). When recorded with electrodes located in the fourth ventricle during surgery, this potential shows the same intracranial spatiotemporal distribution as the P14 component of median nerve SEPs (189). Therefore, the P30 potential is likely to be generated in the brainstem (177,178) and can be viewed as the homolog, for the lower limb, of the far-field P14 recorded on the scalp after median nerve stimulation.

Recording of P30 is made easier if the scalp electrode is placed in the midfrontal region at Fz or Fpz, where P30 has its maximal amplitude with minimal contamination by subsequent cortical potentials. If electrodes are too posteriorly situated, P30 may be lost in the descending slope of the cortical P39 potential. The most practical reference site for recording of P30 is the nucha at the spinous process of the sixth cervical vertebra (191,192). Since no segmental response is evoked by lower limb stimulation at the cervical cord level, the neck can be considered as virtually inactive at the peaking latency of P30. The ascending volley in spinal somatosensory pathways reaches the cervical level at a latency of about 27 msec but, due to the time dispersion of afferent impulses, no consistent potential is recorded by a skin cervical electrode at this latency. In some normal subjects a scalp far-field P27 potential distinct from P30 is recordable in the frontal region, which reflects the ascending volley in the cervical cord (177,178).

In noncephalic or earlobe reference recordings a small N33 negativity (187,188) may precede the P39 potential. This negativity is widely distributed over the scalp; part of it is also picked up at the earlobe so that it is reduced in earlobe reference recordings. Because some of the cortical potentials are picked up at the earlobe contralateral to stimulation, it is preferable to use the earlobe ipsilateral to stimulation as the reference site for the recording of cortical tibial nerve SEPs (194). N33 is considered as the homolog of the N18 upper limb SEP and is likely to have a similar origin in the lower brainstem (see “N18 scalp negativity” above).

The earliest cortical potential elicited by the stimulation of tibial nerve at the ankle is a positive potential usually labeled P37, P39, or P40 for it peaks at a mean latency of 37 to 40 msec in normal subjects (see Ref. 195, for a review). In what follows this
potential is labeled P39. This potential peaks 6 to 7 msec earlier after stimulation of the tibial nerve at the popliteal fossa, and about 3 msec later after stimulation of the sural nerve at the ankle. In normal subjects P39 is recordable on the vertex and can be reliably obtained midway between Pz and Cz using a scalp to earlobe montage. Frequently the maximum of P39, although close to the midline, is slightly shifted on the side of the scalp ipsilateral to the stimulation (196). The most likely reason for this paradoxical lateralization is that the somatotopic representation of the lower limb in the somatosensory SI, and in particular that of the foot, is situated at the inner aspect of the hemisphere. This paradoxical scalp distribution of P39 is not observed when the proximal leg is stimulated, for instance after stimulation of the femoral nerve or lateral femoral cutaneous nerve (197,198). SEP mapping in normals often shows a dipolar field distribution for P39, with a maximum of positivity ipsilateral to the stimulation in the parietal region and a maximum of negativity (N39) in the contralateral frontocentral region (44,187,191,196,199). Magnetic fields evoked by stimulation of the lower limb also show this distribution (42,200). The negative N39 maximum of this dipole is not consistently obtained, probably because of intersubject differences in the orientation of the leg area in SI with respect to the scalp surface, so that one can envisage all intermediates between a vertical radial dipole with a positive maximum on the vertex, and a nearly tangential one with a P39 and N39 culminating, respectively, in the ipsi- and contralateral parietal regions. As for median nerve SEPs, the question is still debated whether all of these potentials reflect the activity of the lower limb area in SI or whether some could be generated in other cortical areas. Another question is to determine whether the P39-N39 dipolar field has a single source in SI tangent to the scalp (191,196,199,201,202), or several (184,203, 204 and 205). Studies on the effect of the frequency rate on cortical SEP amplitude favor the latter hypothesis by showing the following:

Studies of the effects of movement on cortical tibial SEPs have shown that P39, but not N39, is gated during voluntary movement of the stimulated foot (128,208,209). During tonic contraction this effect is proportional to the level of muscle strength maintained by the subjects. These findings have been interpreted in different ways according to the different source models. Tinazzi et al. (210) proposed that P39 could be generated in the motor cortex, while N39 could be the equivalent of the median nerve N20 and generated in area 3b in the SI area. Valeriani et al. (128,208) considered that only the radial central source of the vertex P39, possibly located rostral to the central sulcus, is sensitive to voluntary movements.

P39 is followed by N50 and P60 components, these three waves forming the “W” profile (P39-N50-P60) of the cortical response recorded in the centroparietal region. This waveform has been reported in all studies on tibial nerve SEPs (see Ref. 194, for a review). Electrical stimulation of the L5 and S1 dermatomes also evokes a consistent W-shaped response on the scalp with a first positivity at about 50 msec after stimulation (211). The N50 and P60 potentials culminate at the vertex (Cz) and do not show the same clear dipolar distribution as the P39-N39 response on the scalp, thus suggesting that the W-shaped P39-N50-P60 waveform could be generated by several sources with distinct orientations (187,188). There is no clinical study demonstrating that the N50 or P60 potential can be selectively affected by a single focal lesion, and these two potentials show the same attenuation as P39 during active movement (209). Conversely it is not exceptional in demyelinating diseases to see a single persisting positivity peaking at the P60 latency while the P39 potential cannot be identified.