Location of stimulation‐evoked changes in naming in lateral perisylvian cortex of the left language‐dominant hemisphere in 24‐yr‐old female. Drawing traced from photograph made at operation of exposed cortex, the extent of which is indicated by the crosshatching in the smaller upper drawing. Each stimulated site represented by a rectangle: filled, stimulation‐evoked naming error in all 3 samples; vertical stripes, stimulation‐evoked naming errors in 2 of 3 or 3 of 4 samples; dot, stimulation‐evoked naming error in 1 of 3 or 1 of 4 samples. Arrows identify adjacent sites (distance ≤5mm) on continuous gyri, one site with repeated naming errors, other with no errors at all. Stimulation delivered with 4‐s trains of 60 Hz, 2.5‐ms total duration biphasic pulses of 8 mA between pulse peaks through bipolar electrodes placed 5 mm apart. No errors were made during naming in the absence of stimulation. M and S, Rolandic cortex, as identified by stimulation‐evoked motor and sensory responses at these sites.

Individual variability in location of changes in naming evoked by stimulation of lateral perisylvian cortex of the left language‐dominant hemisphere, based on results from 21 cases. Stimulation sites in each case were aligned by their relation to the motor cortex and to the end of the Sylvian fissure and then assigned to 1 of 8 arbitrary zones identified by dashed lines. One or more samples in each 8 zones was found in 15–19 patients. Multiple samples in 1 zone from a single patient were combined. Circle graphs, proportion of patients with samples in that zone who repeatedly demonstrated significant naming errors; filled area, proportion of cases with anomia; stippled area, proportion with complete arrest of speech as result of stimulation. Smaller proportion of anomic patients indicated by white bar in the anterior superior temporal zone showed naming changes at anterior sites in that zone, immediately below motor cortex.

Patterns of localization of 6 language‐related behaviors in perisylvian cortex of left dominant hemisphere. Stimulation mapping at 9 cortical sites used trains of 60‐Hz, 2.5‐ms total duration biphasic square‐wave pulses of 3 mA between pulse peaks. Symbols at each site indicate performance during different language‐related behaviors. Filled symbols used when performance differed from control (P < 0.05) at given site for given behavior; square at top of each site, performance during naming in Greek, patient's second language; circle immediately below square, performance on naming of same object pictures in English, patient's first language; circles in middle row, performance in reading simple sentences (left) and during short‐term memory task (right); circles in bottom row, performance on measure of phonemic identification (left) and mimicry of orofacial movement (right). Following types of errors were found: all naming errors were anomia; reading errors were arrests; short‐term memory errors varied at different sites: I, errors with stimulation during input to memory; S, memory errors with stimulation during distraction (reading task); O, errors with stimulation during retrieval. Open arrows indicate 2 sites where ability to mimic sequential orofacial movements as well as ability to identify phonemes were altered. Repetitive orofacial movements were not altered at any of the tested sites but the X symbols represent sites where face movements were evoked as part of the initial identification of Rolandic cortex. Large broken circles identify 4 sites with short‐term verbal memory changes, 2 of them independent of any other changes in language‐related behaviors. Separation of sites where naming is altered in both languages is also evident. Patient was 30‐yr‐old woman with verbal IQ of 103. She suffered from temporal lobe seizures since she was 8, after several grand mal seizures at age 4, and was treated with phenytoin at time of craniotomy. Epileptic focus was in temporal tip and did not extend to any of stimulation mapping sites. Control trial error rates were 0% for naming in English, reading, short‐term memory, and mimicry of orofacial movements; 3.6% for naming in Greek; and 3.7% for phoneme identification.

Model of language organization in perisylvian cortex of the dominant hemisphere derived from mapping of naming, reading, verbal memory, phoneme identification, and orofacial mimicry at 118 sites in 14 patients. Maps from individual patients were aligned in relation to motor cortex and the end of Sylvian fissure and then grouped into 16 zones bordered by dashed lines. Within each zone: numbers, number of patients with at least 1 site sampled (most showed only 1 site/zone); star, change in some language function evoked at all sites; small filled circle, zones where stimulation of ∼50% of sites produced no changes in any language function. Upper circle contains symbols defining whether naming and/or reading changes occurred at ∼50% of sites in that zone; horizontal line, reading changes; vertical line, naming changes; heavier lines, zones with naming changes at ∼70% of sites or reading changes at ∼80% of sites. Lower circle contains symbols indicating whether 1 or more divisions of the model were frequently encountered in this zone; filled circle, 50% of sites had features of final motor mechanism; stippling, ∼25% of sites had changes in mimicry of sequences of orofacial movements; slash to left, ∼33% of sites were related to memory; slash to right, ∼50% of sites produced changes in only 1 of the other language functions. In zones with 2 symbols, stimulation at 1 set of sites produced changes in naming, reading, mimicry of orofacial sequences, and phoneme identification but not in memory, whereas stimulation at another set from different patients showed only memory changes. N, zone where majority of evoked changes in naming occurred in the absence of changes in the other measured functions. D, zone where changes in phoneme identification and orofacial movement occurred at separate sites; in other zones where mimicry of orofacial sequences was altered, phoneme changes occurred at most of same sites. G, zones where ∼67% of sites showed reading errors of syntactic type. Site was related to a language function only when change during stimulation differed from non‐stimulation performance at 5% level. Mean age, 28.4 yr (range 17–49 yr); 5 males, 9 females; mean preoperative verbal IQ, 99.4 (range 81–120). Mean stimulating current, 5.3 mA (range 3–8 mA). All stimulations employed biphasic square wave, constant current pulses of 60 Hz, 2.5‐ms total duration, delivered through bipolar electrodes placed 5 mm apart. Mean number of stimulated sites per patient, 8.4 (range 4–16). All tasks measured at 80% of sites; naming, reading, and memory—only at the remainder. Control error rates (mean values followed by range in parentheses): naming, 2.4 (0–6.25); reading, 17.2 (0–41.7); verbal memory, 12.7 (0–35.7); phoneme identification, 13.66 (0–47.6); single orofacial movements, 2.2 (0–11.1); sequential orofacial movements, 4.1 (0–20).

Localization of electrical stimulation‐evoked language changes in thalamus. Circles identify 48 electrode locations in 44 patients, 27 in the right and 21 in the left brain. Open circles, no evoked changes in language as measured by object naming; n, arrest of speech; filled circles, altered naming with intact speech: P, sites where perseverations were evoked; R, sites where same wrong word was frequently used when naming was required during stimulation above threshold current. All other naming changes consisted of anomia. A, area of left anterior superior pulvinar, where ideational language errors of anomic type were evoked in another patient series (cf. ref. 114). Note absence of ideational language errors accompanying stimulation of right thalamus. Coronal sections of thalamus extend posteriorly from anterior commissure at 5‐mm intervals. Mean location of thalamic structures and range of variation of lateral thalamic border was distributed about the same commissural reference also used as reference to localize the electrodes.

Electrocorticographic (ECoG) changes during naming recorded from language‐dominant hemisphere of a 23‐yr‐old male with verbal IQ of 113, suffering from anterior temporal lobe epileptic focus and adult onset seizures. Brain maps show electrode locations. Symbols in circles indicate changes in naming, reading, verbal memory, orofacial movement, and phoneme identification evoked by electrical stimulation: AR, arrest of speech; Mt, spontaneous face motor responses identifying motor cortex; S, sequential (but not repeated single) orofacial movements; P, phoneme identification; R, reading; Mm, verbal memory; N, naming; dash indicates no changes in any measured function. Left column (sN) contains 2 averages of 16 samples of ECoG recorded from each site, starting 400 ms before and lasting until 1,600 ms after 100‐ms presentation of an object picture. Patient was instructed to name pictures to himself and report each name aloud 4 s later at a cue; i.e., the task required silent naming and retention of name in memory. Right column (N with a slash) gives similar ECoG averages recorded during presentation of same object pictures in task requiring reception, matching, and retention of a spatial feature on those pictures. Sites 1 and 9 were identified by stimulation mapping as essential to naming; Each ECoG average obtained at site 1 during silent naming demonstrated a slow potential (SP) beginning 100 ms after presentation of item to be named and lasting ∼1 s (black arrow). Amplitude of potential was quantified for individual tracings by measuring its peak against base line of average activity taken 400 ms prior to presentation of the object picture. Potential increased significantly (P < 0.05) during silent naming and resulting values were significantly greater than values accompanying spatial matching task (comparison a). Averaged ECoG at site 9 demonstrated desynchronization with loss of activity in 7‐ to 12‐Hz range during silent naming. Change was assessed quantitatively on individual tracings by measuring spectral density in 7‐ to 12‐Hz range in successive 0.5‐s intervals. In the 700‐ to 1,200‐ms time segment after presentation of object picture, 7‐ to 12‐Hz activity was significantly reduced compared to 0.5‐s epoch preceding that presentation. During same 700‐ to 1,200‐ms epoch relative rank of site 9 also decreased significantly compared to other temporoparietal recording sites (sites 3–10). When same epoch of the ECoG recorded during the spatial matching task was analyzed, the 7‐ to 12‐Hz spectral density did not differ from that obtained during silent naming but relative rank of the naming site was significantly greater during spatial matching than during silent naming (comparison c). Findings suggest that SP and desynchronization of the ECoG changes were anatomically and behaviorally specific to naming in this patient. Site 8 was related to verbal memory by stimulation. The ECoG recorded at this site during silent naming task (also containing a verbal memory component) demonstrated a negative SP of unique configuration that began ∼200 ms after presentation, peaked at 400 ms, and returned to base line (open arrow). Quantitative assessment of this potential on individual samples was carried out by comparing the peak negative amplitude at 400 ms to 200‐ to 600‐ms base line. Results demonstrated that during silent naming a significant change took place at site 8 without affecting the surrounding sites 3–10. Results were also different from data recorded at site 8 during spatial task that did not have verbal memory component (comparison b); the potential, then, may be related to verbal memory.

Extracellular single‐neuron recordings taken from human temporal lobe lateral cortex during listening and repeating single words (upper tracing of each pair); each tracing includes the superimposed records from 3—10 trials with concurrent audio records (lower tracing). A: activity recorded from right nondominant superior temporal gyrus during listening to 10 multisyllable words; left, marker pulse in audio channel; activity on center and right reflects the presentation of each syllable of the words. Note that concurrent single‐neuron activity increases markedly with onset of 2nd syllable. B: activity recorded from right nondominant superior temporal gyrus of another patient during listening and repeating 3 words. Left, activity in audio channel reflects presentation of words; right, activity identifies patient's repeating the word aloud. Single‐neuron activity increases during word repetition, after delay of ∼0.5 s. C: activity recorded from left dominant middle temporal gyrus of a 3rd patient during listening and repeating 4 words. Note below 2 tracings the averaged ECoG simultaneously recorded from surface at site of microelectrode penetration. Activity in this unit was inhibited during word repetition in association with a positive wave in the surface ECoG.

Tracings and histograms of extracellular neuron activity recorded during a recent verbal memory measurement at a left inferior temporal gyrus site related to recent verbal memory by electrical stimulation mapping. Each trial of the recent verbal memory measurement consists of four 6‐s parts: a, silent naming of an object picture, input to memory; b silent reading of 2 words; c, silent reading of sentence with a blank filled in aloud with 1 of 2 words from b; d, a visual cue for retrieval of name of object pictured in a, stored over the distractors provided by b and c. Top, tracings of single‐neuron activity on 1 trial of this measurement. In histograms: 1, average firing activity over 6 trials, divided into 2‐s epochs for each of a to d parts; 2 and 3, activity from same site during 4 control measurements; 2, presentations of same slides as in a, but with the instruction to silently detect spatial feature that must be matched on a subsequent slide. This control measurement requires neither naming nor recent verbal memory; 3, undelayed naming of same object picture: 3s, naming silently; 3o, naming overtly. This control measurement requires naming but makes no demands on recent verbal memory; 3b, undelayed silent reading of same words as in 1b. Differences between recent‐memory measurements (1a‐d) and control conditions (2–3) that reach statistical significance (P < 0.05) include comparison between 1a and both 2a or 3s in the 2‐ to 6‐s time period, indicating increased firing when naming is the input to recent verbal memory, compared to use of same picture in spatial task, or in naming without memory; comparison between the first 2 s of part 1b and control measure 3b indicating that increased firing lasts into beginning of distractor task but ceases during later epochs when memory must be stored; comparison between all time epochs of 1d and 3o where same words are produced, difference being that in 1d the words require retrieval from verbal memory to visual cue, but in 3o they are repeated as overt naming. Differences within histograms of recent‐verbal‐memory tasks (1) were determined by comparison to last epoch of b portion of task, when there was no measurable difference in neuron activity between memory and control. Last portion of b epoch of task comes as close as possible to behaviorally neutral condition. First and 2nd epoch of part a of task, last epoch of c, and middle epoch of d showed statistically significant increases in single‐neuron firing. These findings from site related to recent verbal memory were interpreted as showing prolonged increase in single‐neuron firing after input and retrieval of items from recent verbal memory, but not throughout memory storage.