Localization and characterization of speech arrest during transcranial magnetic stimulation

Clinical Neurophysiology 110 (1999) 1073±1079

Localization and characterization of speech arrest during transcranial magnetic stimulation Charles M. Epstein a, d,*, Kimford J. Meador c, David W. Loring c, Randall J. Wright a, Joseph D. Weissman a, Scott Sheppard b, James J. Lah a, Frank Puhalovich a, Luis Gaitan a, Kent R. Davey e b

a Department of Neurology, Emory University School of Medicine, Atlanta, GA 30322, USA Frederick Philips Magnetic Resonance Research Center, Emory University School of Medicine, Atlanta, GA 30322, USA c Department of Neurology, Medical College of Georgia, Augusta, GA 30912, USA d Rehabilitation Research and Development Center, Atlanta Veterans Affairs Medical Center, Atlanta, GA 30033, USA e Neotonus, Inc., Marietta, GA 30060, USA

Accepted 16 February 1999

Abstract Objective: To determine the anatomic and physiologic localization of speech arrest induced by repetitive transcranial magnetic stimulation (rTMS), and to examine the relationship of speech arrest to language function. Methods: Ten normal, right-handed volunteers were tested in a battery of language tasks during rTMS. Four underwent mapping of speech arrest on a 1 cm grid over the left frontal region. Compound motor action potentials from the right face and hand were mapped onto the same grid. Mean positions for speech arrest and muscle activation were identi®ed in two subjects on 3-dimensional MRI. Results: All subjects had lateralized arrest of spontaneous speech and reading aloud during rTMS over the left posterior-inferior frontal region. Writing, comprehension, repetition, naming, oral praxis, and singing were relatively spared (P , :05). Stimulation on the right during singing abolished melody in two subjects, but minimally affected speech production. The area of speech arrest overlay the caudal portion of the left precentral gyrus, congruous with the region where stimulation produced movement of the right face. Conclusions: The site of magnetic speech arrest appears to be the facial motor cortex. Its characteristics differ from those of classic aphasias, and include a prominent dissociation among different types of speech output. q 1999 Elsevier Science Ireland Ltd. All rights reserved. Keywords: Transcranial magnetic stimulation; Speech; Language; Frontal lobe

1. Introduction Active localization of language function has traditionally been possible only with invasive procedures. The dominant hemisphere can be determined using the intracarotid amobarbital procedure or Wada test (Wada and Rasmussen, 1960). Cortical areas critical to language are mapped using direct electrical cortical stimulation (DCS) in the operating room (Pen®eld and Rasmussen, 1950) or extra-operatively through electrode grids implanted in the subdural space (Lesser et al., 1987). The Wada test and electrocorticography have contributed greatly to current ideas about language organization. Because of their invasiveness and potential * Corresponding author. Department of Neurology, 1365 Clifton Road, N.E., Atlanta, GA 30322, USA; Tel.: 1 1-404-778-3633; fax: 1 1-404778-3767.

morbidity, these techniques are con®ned almost entirely to patients undergoing surgery for intractable epilepsy. In the past decade positron emission tomography and functional magnetic resonance imaging (MRI) have shown exciting results for language localization (Hinke et al., 1993; Petrides et al., 1995; Binder et al., 1996). However, metabolic changes may be present in brain areas that are not essential for a speci®c language behavior (Ojemann, 1991). At least 4 groups have reported lateralized speech arrest using repetitive transcranial magnetic brain stimulation (rTMS). Sensitivity was 100% in two small series (Pascual-Leone et al., 1991; Epstein et al., 1996), but only 5067% in others (Jennum et al., 1994; Michelucci et al., 1994). Speech interruption was more prominent over the left hemisphere, and, when assessed in epilepsy patients, showed a high correlation with the Wada test. Most of these studies used large circular magnetic coils, along with stimulus para-

1388-2457/99/$ - see front matter q 1999 Elsevier Science Ireland Ltd. All rights reserved. PII: S 1388-245 7(99)00047-4

CLINPH 98114

1074

C.M. Epstein et al. / Clinical Neurophysiology 110 (1999) 1073±1079

meters that may carry a risk of inducing seizures (PascualLeone et al., 1993). Extensive language testing was not performed. Thus, the initial rTMS techniques were not optimal for detailed localization or for studies involving normal subjects. We recently described studies of rTMS speech arrest using a more powerful and focal coil. Greater power allowed the production of lateralized speech arrest with a repetition rate as low as 2±4 Hz, and a combination of stimulus parameters that comply with recent safety recommendations (Epstein et al., 1996). The slower stimulation rate reduces the discomfort of rTMS despite the greater intensity of individual pulses. Thus, this technique is feasible for detailed investigation of magnetic speech arrest in normal individuals. The ®rst such studies are described below. 2. Materials and methods 2.1. Subjects Subjects were 10 naturally right-handed adults, aged 21± 48 (average 35.7). Seven were male and 3 were female. All gave informed consent to a protocol approved by the Human Investigations Committee of the Emory University School of Medicine. Four were among the authors of this paper; the others were volunteers who were compensated for their participation. The volunteers were naive within the limits of informed consent; they knew that the magnetic stimulator was expected to affect speech output, that it would be applied over different brain areas, and that they would undergo a battery of language tests. Potential subjects were administered a 14-item handedness inventory (Raczkowski and Kalat, 1974), and were further tested for lateralization of speech arrest during counting, as described below. All 10 volunteers who scored 14/14 for right handedness and had lower-threshold arrest of speech over the left hemisphere were chosen for this study. 2.2. Magnetic mapping of speech arrest This portion of the study included 4 subjects, who were seated comfortably and unrestrained. The head was covered with a latex swim cap, which simpli®ed position measurements over a large scalp area that included up to 100 possible sites of stimulation. Any redundant latex folds were taped down, and the position of the cap was labelled, using as landmarks, the inion, nasion, earlobes, and vertex. One centimeter grid lines were drawn over the posterior frontal region and labelled numerically. Relaxed motor threshold was determined as previously described (Epstein et al., 1996), using the dominant ®rst dorsal interosseous or abductor pollicis brevis to represent the hand. With this technique, threshold is de®ned as the lowest intensity of stimulation that produces compound motor action potentials (CMAPs) of 50 mV or greater, on 5 of 10 trials (Pascual-

Leone et al., 1993); consequently averaged CMAPs are expected to be non-zero at threshold. Mapping of CMAPs and speech arrest was performed using a prototype iron-core stimulation coil, with the induced electric ®eld maximum beneath the center point of the device (Epstein et al., 1996). This coil was more powerful and more focal than available commercial devices. Each pulse was biphasic with a duration of 180 microseconds, and the magnetic ®eld distribution was comparable to that from a ®gure-8 coil with outer dimensions of 5 by 10 cm. A small port through the middle allowed precise marking and positioning. The coil was oriented so that the induced electric ®eld was aligned anterior±posterior. With the right hand relaxed, we averaged 8 responses to left hemisphere stimulation at each site, using a rate of 1 Hz. Testing was extended in all directions on the grid until a 2 cm rim of absent responses completely surrounded the area of activation. Mapping was done in the same manner using the right orbicularis oris, but facilitation was used if no response could be obtained during relaxation at stimulator outputs up to 20% greater than hand motor threshold. Speech interruption was tested with the same coil orientation and with stimulus intensity set initially to hand motor threshold. Subjects were instructed to count upwards briskly, but distinctly. After a few seconds of counting, magnetic stimulation was begun at 4 Hz while the coil was moved across the presumed dominant fronto-temporal region. If no speech effect was obtained, intensity was increased by 10% of motor threshold up to a maximum of 150% for 5 s, according to preliminary recommendations for safety in rTMS (Pascual-Leone et al., 1993). When a region of complete speech interruption was identi®ed, the area of effect was estimated through movement of the stimulator; the apparent center point was then identi®ed and marked as the primary test position. We next applied the coil to the homologous position on the right side, using the same intensity. At that point and in a surrounding circle of 4 cm radius, an attempt was made to obtain speech arrest during counting and reading. The degree of speech interruption at each scalp position was rated concordantly by two observers as complete, moderate, slight, or absent. This part of the protocol required about 2 h per subject. 2.3. Language testing In all 10 subjects, the lowest-intensity site and lowest stimulator output for complete speech arrest were determined as described above. At this intensity, the following language tasks were carried out during stimulus trains of 3± 5 s duration: ² reading unfamiliar material aloud; ² reading silently during stimulation and afterwards describing content; ² spontaneously describing the events of the `cookie theft picture' (Goodglass and Kaplan, 1983). For this task subjects ®rst examined the picture, mentally prepared a

C.M. Epstein et al. / Clinical Neurophysiology 110 (1999) 1073±1079

1075

² singing lyrics to a familiar song; ² tests of oral praxis, including tapping the upper teeth with the tongue, licking the lips from side-to-side, alternating lip puckering and blowing, repeating the syllable `pa-papa-pa,' and elongating the single syllable `paaaaaa.'.

Fig. 1. Averaged compound motor action potentials from the right hand and face of one subject. Note different time scales.

² ² ² ²

few sentences describing the events depicted, and then continued viewing the picture while attempting to speak those sentences during stimulation; hearing and obeying two-step commands with inverted syntax, both during stimulation; visual confrontation naming of familiar objects; writing numbers from `one' upwards, using the right hand; repetition of two brief phrases, including `no ifs, ands, or buts', with stimulation both while listening and speaking;

Fig. 2. Magnetic bubble map of the left lateral frontal region in subject 1, showing responses from the right ®rst dorsal interosseous (hand) and right orbicularis oris (face), plus the extent of speech arrest at each site. The mean positions for hand, face and speech effects are indicated by H, F and S, respectively. Stimulus intensities are given relative to resting motor threshold in the hand. Areas of the bubbles correspond to normalized amplitudes of compound motor action potentials and to estimated degree of interference with speech. Positions of bubbles for hand and speech are slightly displaced vertically to reduce overlap, whereas all mean positions are exact with respect to the grid.

As with counting, most tasks were begun several seconds before the onset of magnetic stimulation. However, the twostep commands, objects to be named, and brief phrases to be repeated were presented simultaneously with the onset of the stimulus trains. Writing and visual confrontation naming were then repeated with the stimulator placed 2 cm anterior to the primary test position on the left (a location meant, on the basis of the mapping results, to approximate Broca's area.) In addition, singing was repeated during stimulation of the homologous area over the right hemisphere. Interruption of spontaneous counting was retested at intervals during the other language protocols to verify continued effect. Performance on each task was scored at the time of testing by two observers using a range from 0±5, with zero representing no detectable interference by rTMS. The paired scores, which never differed by more than one point, were averaged to give the ®nal rating. A delay of 20 s or more was always present between stimulus trains. Subjects were asked to rate the discomfort of 4 Hz stimulation on a scale of 0±10. This portion of the protocol required about 1/2 h per subject. Results of the language battery were compared non-parametrically using the Wilcoxon sign test. 2.4. Localization of magnetic effects During construction of a physiological map the average CMAPs representing each muscle were normalized to a maximum of one. Complete speech arrest was arbitrarily assigned a magnitude of 1.0, moderate speech interruption 0.5, and slight speech interruption 0.25. Bubble charts were plotted with the area of each bubble corresponding to the magnitude of the response at that site. For hand and face muscles and for speech arrest, we calculated two-dimensional mean positions (centers of gravity) on the 1 cm grid (Miranda et al., 1997). Three-dimensional MRIs were obtained for two subjects. The two-dimensional mean positions were marked on the original swim caps, which were replaced on the subjects' heads, and realigned to the previous anatomic landmarks. Each mean position was marked with a capsule of vitamin E for identi®cation on MRI. The pulse sequence was a fast ®eld echo with a 230 mm ®eld of view, TR ˆ 33 ms, TE ˆ 11 ms, tipangle ˆ 358 and resolution 256 £ 256 £ 70. Images were processed using an EasyVision workstation (Philips Medical Systems, Shelton, CT), with user-assisted segmentation and 3-dimensional surface reconstruction. In the ®nal images, position labels were aligned over the centroids of the vitamin E capsules.

1076

C.M. Epstein et al. / Clinical Neurophysiology 110 (1999) 1073±1079

Fig. 3. Magnetic bubble map in subject 2.

Fig. 5. Magnetic bubble map in subject 4.

3. Results

3.2. Language testing

3.1. Localization of speech arrest Motor maps were constructed for the right hand using ®rst dorsal interosseous in 3 subjects and abductor pollicis brevis in one. Two subjects had facial CMAPs obtainable from the orbicularis oris at rest (Fig. 1). The other two maps of orbicularis oris required facilitation through gentle pursing of the lips, which lowers the motor threshold. Left hemisphere sites for activation of muscles in the right face and hand are shown on the grid positions in Figs. 2±5, together with the locations and extent of speech arrest and the mean positions for all 3 effects. Although testing was performed with different relative intensities of rTMS, the region where stimulation produced speech arrest was always congruous with the area which gave motor responses from orbicularis oris. Across all 4 mapping studies, the grand mean position for speech arrest lay 0.6 cm superior and 0.3 cm posterior to that for orbicularis oris. On the 3-dimensional MRI images, the mean positions for the ®rst dorsal interosseous, orbicularis oris, and speech arrest lay close to the central sulcus (Figs. 6 and 7).

Complete speech arrest was obtained over the left posterior-lateral frontal lobe during counting in all 10 subjects. Repetition of the counting task produced the same results at the same intensity throughout the course of testing. The delay from onset of magnetic stimulation to total arrest of speech was 1±3 s. In 3 subjects, speech interruption consisted of total mutism, while the others were able to generate simple sounds such as `uh-uh-uh' that were completely unintelligible. Complete speech arrest occurred, on average, at 116% of hand motor threshold (range 100± 137%), the maximum being within recent, modi®ed safety recommendations for class 3 studies of rTMS (Wassermann, 1998). The average pain rating for 4 Hz rTMS at this intensity was 4.2 (range 2±7). In 8 subjects, counting was entirely normal on right-sided stimulation at the same intensity. The others had mild dysarthria but not speech arrest with stimulation on the right. Subjective reactions to speech arrest using this technique have been described previously (Epstein et al., 1996). Average scores and statistical comparisons for the language-related tasks are given in Table 1. The differences

Fig. 4. Magnetic bubble map in subject 3.

Fig. 6. MRI of subject 3, corresponding to Fig. 4, with mean stimulus positions for the right ®rst dorsal interosseous, orbicularis oris and speech arrest projected onto the cortical surface. Arrows indicate the central sulcus.

C.M. Epstein et al. / Clinical Neurophysiology 110 (1999) 1073±1079

1077

subjects. Because the higher intensity of stimulation required on the right was more painful, and in some cases would have exceeded safety recommendations available at the time of study (Pascual-Leone et al., 1993), comprehensive right-sided stimulation was omitted in the remaining subjects. 4. Discussion

Fig. 7. MRI of subject 4, corresponding to Fig. 5.

in task performance followed the same pattern in both the naive volunteers and the 4 authors who participated. At the optimal site for interference with counting, subjects also had great dif®culty speaking spontaneous sentences and reading aloud; but they performed signi®cantly better at repetition, singing, tests of oral praxis, and visual confrontation naming. 1 Naming and repetition of spoken phrases were often accomplished with slowing of responses, variable dysarthria, and occasional stuttering or explosive vocalizations. Aphasic errors were rare, and similar to those noted by others (Pascual-Leone et al., 1991). Interference with naming was always less prominent during stimulation anterior to the primary test position, in the expected region of Broca's area. Writing numerals was intact in the right hand for 8 subjects at both left frontal sites of stimulation. One subject was unable, on repeated attempts, to generate legible numerals beyond 1±5, and another had slow irregular handwriting in the absence of stimulation-related movement. Both of these effects occurred at the primary test site. Reading and obeying two-step commands with inverted syntax was normal in 9 subjects, and moderately impaired in one. Ability to read silently during stimulation and then describe the material was normal in 7 subjects, and judged moderately impaired in 3. Oral praxis was minimally impaired, except in one subject, who had moderate dif®culty with tongue movement. Of all praxis tests repetition of the simple syllable `pa' was most dif®cult across subjects, and is compared with other tasks in Table 1. Singing was consistently much easier than spontaneous speech, with slight to moderate slowing or dysarthria, but with preservation of melody. However, in two of 8 subjects so tested, stimulation over the right hemisphere produced ¯attening and loss of melody that was apparent to both the subject and the observers. Right hemisphere rTMS had no detectable effect on reading, naming, or repetition in 4 1

In preliminary analysis, counting appeared signi®cantly more dif®cult than all other tasks. However, because complete arrest of counting was always de®ned as `5', it was subject to a ceiling effect. Estimates of reading aloud were therefore considered less biased, and preferred for the statistical comparisons listed in Table 1.

In a group of right-handed volunteers, chosen for a strong probability of left hemisphere dominance, rTMS over the left inferior frontal area produced speci®c impairments of verbal output. Some modalities of speech were affected profoundly, but others moderately or not at all. Magnetic interference dramatically impaired self-generated spontaneous speech. It had less effect on repetition, confrontation naming, singing, comprehension, and oral praxis; writing was usually spared. The site of action was congruous with the region of facial motor responses, rather than anterior to the motor strip as might be expected from classic models of language organization. Despite the consistent lateralization of effects, in this study rTMS impaired primarily motor speech rather than the generation of language. Flitman et al. (1998) have shown that independent of speech arrest, linguistic processing is disrupted by rTMS over anterior and posterior sites in the left hemisphere but not the right. However, the magnitude of this effect was relatively modest, and its demonstration required hundreds of rTMS trains in each subject. Many features of magnetic speech arrest resemble those of the articulatory disorders variously described as pure motor aphasia, cortical anarthria, speech apraxia, phonetic disintegration, and aphemia. Such cases have been described with subcortical lesions of the lateral frontal region, but when cortical lesions are responsible they are found in the lower motor strip and Rolandic operculum (Lecours and Lhermitte, 1976; Tonkonogy and Goodglass, 1981). Clinical ®ndings include marked slowing with irregularly explosive speech output, stuttering, and preservation of grammar. Writing is unaffected. However, patients with aphemia are dysarthric in all speech production tasks. In this respect, the speech disorder produced by rTMS differs from both aphemia and the classical aphasias, being less profound when the generation of phonemes is cued by speech reception, melody, or the presentation of familiar visual objects. The ease of naming objects may be due, in part, simply to the brevity of the required response. But repetition during rTMS is relatively spared as well, suggesting additional factors besides the length of utterance. The best-known technique for physiological mapping of language function is DCS. Speech arrest may be obtained with DCS over extensive areas of both hemispheres (Pen®eld and Rasmussen, 1950). Other types of negative motor responses can be produced from either hemisphere when DCS is applied just rostral to the primary motor

1078

C.M. Epstein et al. / Clinical Neurophysiology 110 (1999) 1073±1079

Table 1 Mean scores for language-related tasks. `0' is normal function, `5' is complete interference; P values represent reading aloud versus the task listed in that column Subject

Age

Reading aloud

Generating sentences

Singing left

Naming

Repeating

Reading comprehension

Singing right

Writing

Inverse syntax

Praxis (pa-pa-pa)

1 2 3 4 5 6 7 8 9 10 Mean Pa

34 48 32 39 33 34 35 47 34 21 35.7 ±

2 3 4 4 4 5 4 5 4 5 4.0 ±

3 2 4 3 4 2 4 5 5 5 3.7 0.500

1 1 ± 3 2 1 2 5 2 4 2.3 0.012

4 1 1 1 2 2 2 5 2 2 2.2 0.008

1 1 3 1 2 1 2 2 2 5 2.0 0.008

0 0 3 0 3 0 3 0 0 0 0.90 0.005

0 ± ± 1 2 0 1 1 0 2 0.88 0.012

0 0 0 0 0 4 0 1 0 0 0.50 0.005

0 0 0 3 0 0 0 0 0 0 0.30 0.005

3 1 0 3 0 1 1 5 0 3 1.7 0.015

a

Signi®cance from Wilcoxon sign test.

cortex, in a region that overlaps with the classical Broca's area on the left (LuÈders et al., 1987). However, the most frequent site for interruption of speech is the facial portion of primary motor cortex, near the junction of the Sylvian and Rolandic ®ssures (Pen®eld and Rasmussen, 1950). Dominant hemisphere sites essential to visual confrontation naming are also widely distributed, but appear to be found most often in the frontal perisylvian area, immediately adjacent to the central sulcus (Ojemann, 1991). Thus the region implicated in magnetic speech arrest over the dominant hemisphere also appears to be the most common site of speech impairment evoked by DCS during electrocorticography, and is associated pathologically with articulatory speech disorders (Tonkonogy and Goodglass, 1981). In contrast to DCS, rTMS appears to block speech generation at only one location, and to do so with greater consistency. However, rTMS has little effect on visual confrontation naming in the area where it has the most obvious effects on spontaneous speech. rTMS at 4 Hz is safer and less uncomfortable than at faster rates (Epstein et al., 1996), but subjective pain estimates up to 7/10 are not trivial. Thus, the number of language-mapping tasks that could be carried out in each subject was ®nite. It is possible that stronger magnetic stimulation at different sites might produce more prominent effects on other language functions, comparable to those of DCS. However, investigations using more powerful stimulation could be quite dif®cult, given the limitations of pain, seizure hazard, and an increasingly large area of activation as power output is raised. All stimulation parameters in this study complied with the most recent and conservative recommendations for safety in rTMS (Chen et al., 1997; Wassermann, 1998); the latter may now represent a more important limitation than available technology. Nonetheless, magnetic mapping of the cerebral cortex has advantages that go beyond its ease of use. One of these, obviously, is the ability to study both hemispheres of the normal brain. Another is the robustness of effects. DCS in

conscious subjects fails to activate any hand movement in as many as 35%, fails to produce movement of the face or tongue in up to 82%, and occasionally fails to identify areas of speech arrest anywhere in the dominant frontal lobe (Ebeling et al., 1992; Ojemann et al., 1993). These limitations may be due, in part, to sedation and time restrictions during operative stimulation, but are still present in more leisurely studies utilizing chronic subdural grids (LuÈders et al., 1987). Thus, a physiological distinction between different cortical regions cannot always be made during electrocorticography; there are simply not enough sites of activation in a given patient. But appropriate magnetic stimulation will always activate multiple hand muscles in normal subjects, and with the technique used here it has consistently produced inferior frontal speech arrest in subjects surveyed thus far. The reliability of rTMS may relate to a more uniform electric ®eld vector affecting a larger volume of cortex. At present, however, DCS appears more effective at blocking both the generation and comprehension of language. The rapid, precise, coordinated synthesis of multiple lingual-buccal-vocal movements into long sequences of phonemes represents one of the most extraordinary tasks carried out by the human motor system ± which may help explain its vulnerability to rTMS at a rate of only 4 Hz. Ojemann and Mateer (1979) have speculated that specialized language functions arose phylogenetically from a lateralized perisylvian motor control system for sequential orofacial movements. Analysis of magnetic speech arrest supports the current interpretation of language organization as modular, rather than the old concept of a single speech area that carries out multiple functions. But different functional modules may still overlap in the same cortical space. Thus, the ®nal common pathway for the assembly of phonemes may lie within the motor cortex, where it would be dif®cult to bypass using parallel pathways of language processing. rTMS over this area appears to have a selective effect on speech output, especially if self-gener-

C.M. Epstein et al. / Clinical Neurophysiology 110 (1999) 1073±1079

ated or involving long and complex motor sequences. This impediment can be overcome, to varying degrees, by other cognitive cues such as visual images, melody and repetition. Acknowledgements We are indebted to Harold Goodglass, Ph.D., for helpful suggestions and comments. References Binder JR, Swanson SJ, Hammeke TA, Morris GL, Mueller WM, Fischer M, Benbadis S, Frost JA, Rao SM, Haughton VM. Determination of language dominance using functional MRI: a comparison with the Wada test. Neurology 1996;46:978±984. Chen R, Gerloff C, Classen J, Wassermann EM, Hallett M, Cohen LG. Safety of different inter-train intervals for repetitive transcranial magnetic stimulation and recommendations for safe ranges of stimulation parameters. Electroenceph clin Neurophysiol 1997;105:415±421. Ebeling U, Schmid UD, Ying H, Reulen HJ. Safe surgery of lesions near the motor cortex using intra-operative mapping techniques: a report on 50 patients. Acta Neurochir 1992;119:23±28. Epstein CM, Lah JK, Meador K, Weissman JD, Gaitan LE, Dihenia B. Optimum stimulus parameters for lateralized suppression of speech with magnetic brain stimulation. Neurology 1996;47:1590±1593. Flitman SS, Grafman J, Wassermann EM, Cooper V, O'Grady J, PascualLeone A, Hallett M. Linguistic processing during repetitive transcranial magnetic stimulation. Neurology 1998;50:175±181. Goodglass H, Kaplan E. The Assessment Of Aphasia And Related Disorders. Philadelphia: Lea & Febiger, 1983. Hinke RM, Hu X, Stillman AE, Kim SG, Merkle H, Salmi R, Ugurbil K. Functional magnetic resonance imaging of Broca's area during internal speech. NeuroReport 1993;4:675±678. Jennum P, Friberg L, Fuglsang-Frederiksen A, Dam M. Speech localization using repetitive transcranial magnetic stimulation. Neurology 1994;44:269±273. Lecours AR, Lhermitte F. The `pure form' of the phonetic disintegration syndrome (pure anarthria); anatomo-clinical report of a historical case. Brain Lang 1976;3:88±113. Lesser RP, LuÈders H, Klem G, Dinner DS, Morris HH, Hahn JF, Wylie E.

1079

Extraoperative cortical functional localization in patients with epilepsy. J Clin Neurophysiol 1987;4:27±53. LuÈders H, Lesser RP, Morris HH, Dinner DS, Hahn J. Negative motor responses elicited by stimulation of the human cortex. Adv Epileptol 1987;16:229±231. Michelucci R, Valzania F, Passarelli D, Santangelo M, Rizzi R, Buzzi AM, Tempestini A, Tassinari CA. Rapid-rate transcranial magnetic stimulation and hemispheric language dominance: usefulness and safety in epilepsy. Neurology 1994;44:1697±1700. Miranda PC, de Carvalho M, Conceicao I, Luis ML, Ducla-Soares E. A new method for reproducible coil positioning in transcranial magnetic stimulation mapping. Electroenceph clin Neurophysiol 1997;105:116± 123. Ojemann GA. Cortical organization of language. J Neurosci 1991;11:2281± 2287. Ojemann GA, Mateer C. Human language cortex: localization of memory, syntax, and sequential motor-phoneme identi®cation systems. Science 1979;205:1401±1403. Ojemann GA, Sutherling WA, Lesser RP, Dinner DS, Jayakar P, SaintHilaire JM. Cortical stimulation. In: Engel J, editor. Surgical Treatment of the Epilepsies, New York: Raven Press, 1993. pp. 399±414. Pascual-Leone A, Gates JR, Dhuna A. Induction of speech arrest and counting errors with rapid-rate transcranial magnetic stimulation. Neurology 1991;41:697±702. Pascual-Leone A, Houser CM, Reese K, Shotland LI, Grafman J, Sato S, Valls-Sole J, Brasil-Neto JP, Wassermann EM, Cohen LG, Hallett M. Safety of rapid-rate transcranial magnetic stimulation in normal volunteers. Electroenceph clin Neurophysiol 1993;89:120±130. Pen®eld W, Rasmussen T. The Cerebral Cortex Of Man. New York: MacMillan, 1950. Petrides M, Alivisatos B, Evans AC. Functional activation of the human ventrolateral frontal cortex during mnemonic retrieval of verbal information. Proc Natl Acad Sci USA 1995;92:5803±5807. Raczkowski D, Kalat JW. Reliability and validity of some handedness questionnaire items. Neuropsychologia 1974;12:43±47. Tonkonogy J, Goodglass H. Language function, foot of the third frontal gyrus, and Rolandic operculum,. Arch Neurol 1981;38:486±490. Wada J, Rasmussen T. Intracarotid injection of sodium amytal for the lateralization of cerebral speech dominance. J Neurosurg 1960;17:266±282. Wassermann EM. Risk and safety of repetitive transcranial magnetic stimulation: report and suggested guidelines from the International Workshop on the Safety of Repetitive Transcranial Magnetic Stimulation. June 5± 7, Electroenceph clin Neurophysiol 1998;108:1±16.