A. Location of the linear microelectrode array in anteroventral IT is indicated with a yellow arrowhead on a sagittal MRI taken with the electrodes in situ. The probe is in the lateral aspect of the fusiform g., medial bank of lateral occipito-temporal s. (coordinates 38 lateral, 22 posterior, 11 down) (). Patient 1.
B. Potential gradient (PG) recorded between contacts at 150μ centers, during 2 Word Recognition trials (WR). Spontaneous background activity between trials is dominated by 4-5 Hz oscillations (↓) in middle to upper cortical layers (channels 6 to 14, periods with white background). Presentation of new or old words (red and green bars at bottom) suppresses the theta activity (↑), especially from ~300-800ms after word onset (periods with violet background).
C. iERSP plotted as a z-score relative to the pre-stimulus baseline for each frequency and channel, allowing the task-related modulation of theta to be examined and compared to other frequencies. In the new and old plots, the values are averaged across trials and Red indicates increases in CSD power in that frequency, latency and channel from baseline; blue indicates a decrease. In the new-old plots, iERSP is expressed as a t-test between trials with new versus old words. Red indicates more power to new than old words in that frequency, latency and channel; blue indicates less. Data is presented from a Size Judgment (SJ) task to words referring to objects or animals. Similar results were obtained in WR (see ), and a Verb Conjugation task where the subject detects regular past tense morphology (not shown). The initial response is a strong wideband increase in spectral power at ~130-300ms (pink background) in upper and especially middle layers (↓). New and Old are initially similar, but significantly diverge starting ~230ms, when the spectral power increase to the Olds abruptly ends, whereas that to the News continues until ~300ms (↓↓) resulting in significant new/old differences (⇓). During the next phase, lasting until ~800ms post-stimulus onset (violet background), theta activity profoundly decreases compared to baseline (^). The decrease is deeper to new words (^^) resulting in significant new/old differences (Δ). In the third phase, until ~1400ms (yellow background), theta increases in deep layers to old words (↑), resulting in significant old>new differences (⇑). The overall pattern of response as well as the effects of word repetition were similar in all tasks, regardless of whether repetition defined explicit recognition targets, or was incidental to task requirements.
D. Middle layer CSD waveforms from individual trials aligned on stimulus onset and superimposed. In the pre-stimulus period, the CSD is large, and dominated by low frequencies with no discernable phase-relationship (^). This activity is abruptly terminated by a sharp current sink (downward deflections) peaking at ~180ms (▲), evolving into synchronous theta oscillations (⇑⇓) that are more prominent for old (green, 31 trials) than new words (red, 23 trials). The sharp peak is ~65ms duration, corresponding to ~16 Hz. Theta source peaks are visible at ~300, ~500 and ~700ms, after which the theta rapidly loses its phase-locking with the stimulus. Raw waveforms were high pass filtered at 1 Hz. Trials were selected for display that evoked a large amplitude initial sink from 170 to 200ms on ch 12.
E. Laminar profile of theta-generating transmembrane current flows. CSD amplitude is plotted against time to show average waveforms of theta cycles in different cortical layers. These demonstrate that theta is generated by sinks in middle cortical layers, (▲; purple box) accompanied by sources in more superficial (↓; orange box) and (weakly) in deep layers (green box). Over time, these alternate with sources in middle layers (⇓) and sinks (⇑) in superficial and (weakly) deep layers. Separate waveforms are shown for the second through fourth theta cycles after old (green) or new (red) word presentation (these occurred approximately in the range of 300-800ms after word onset), and for the second through fourth theta cycles prior to the stimulus (blue). Amplitude profiles of the three waveforms across cortical layers are indistinguishable, implying that event-related and spontaneous thetas are generated by similar or identical intracolumnar synaptic circuitry. Theta is generated through alternating sources and sinks in middle and superficial cortical layers with accompanying sources in superficial and deep layers.
F. CSD amplitude plotted against cortical depth at the trigger time in Figure 1E. The most prominent features of theta generation are a middle layer sink (▲) with superficial (⇓) and deep sources (↓). No difference can be observed between the amplitude profiles of theta generated after the stimulus to new (red) or old (green) words, versus spontaneous theta in the prestimulus period (blue).
G. Theta phase (from the Hilbert transform) averaged with respect to stimulus onset, demonstrating phase-locking with trial onset. Middle-layer phase shows positive peaks at ~140, 340 and 540ms (⇓) after stimulus onset, i.e., at 5Hz. Phase becomes random after ~800ms. Prestimulus phase is also random and thus averages to zero. Note that the Hilbert phase is a nonlinear measure because it abruptly transitions from –π to +π, and thus may appear slightly different from the ITC measure.
H. Event-related averaged CSD. As reported elsewhere (), the initial response is a sink in layer IV at ~180ms (△). Phase-resetting aligns the CSD waves so that traditional averaging methods reveal an evoked response that includes strong theta frequency oscillations (↓) even though theta power has decreased from the pre-stimulus period.
I. Spectral power, averaged on the peak of the sink in Layer III, increases in a wide band up to ~150Hz. Scale same as C except pixels with zscore less than 3 have been masked as green.

A. Yellow arrow indicates the laminar probe location in a sagittal MRI taken with the probes in situ. Patient 2, perirhinal cortex (PR): laminar tip in the lateral aspect of the parahippocampal g., medial bank of collateral s. (Talairach coordinates 31, −22, −16).
B. Continuous PG recording during explicit memory (WR). Spontaneous background activity between trials (white background) shows large 4-6 Hz oscillations (↓). Presentation of words suppresses theta (↑), especially from ~300-800ms (periods marked with violet backgrounds).
C. iERSP evoked by words in the WR (above) and SJ (below) tasks. The initial response is a strong broadband increase from ~130-300ms (pink background), that is earliest and strongest in putative layer IV (^), but rapidly spreads to superficial and deep layers (Δ). This initial burst is followed by sustained broadband activity until ~700-800ms, especially in middle and deep layers (↓⇓), and theta in upper layers decreases compared to baseline (↑⇑). The later part of the broadband increase is larger to new words, as is the theta decrease (↓↑). Similar iERSP patterns are evoked by words referring to animals versus objects in the SJ task. The initial broadband increase in PR spectral power is stronger to animals (↓) than to objects (⇓), as is the subsequent increase in theta (↑⇑).
D. Recordings from single trials, aligned on stimulus onsets and superimposed. Large theta activity with random phase is visible as the thick cloud of traces in the pre-stimulus period and again more than 1200ms after the stimulus (^). Stimulus onset evokes a sharp deflection peaking at ~140ms in the middle cortical layers (↓). The sharp peak is ~80ms duration, corresponding to ~13 Hz. This is followed by a slower oscillation (⇑⇓) from ~200-400ms (~200ms duration, corresponding to ~5Hz, i.e., theta). Overall activity remains suppressed until ~700ms, then gradually recovers by ~1200ms. During this recovery, theta, although small, appears to be synchronized in the superimposed single sweeps. Data is highpass filtered at 1 Hz, and includes 36 trials of old words (green) and 36 of new (red).
E. CSD averaged with respect to theta phase peaks identified using the Hilbert transform. The dominant CSD during theta in PR is a sink (↑) in middle cortical layers III/IV (purple box), with an accompanying source (↓) in layers I/II (orange box) and V (green box). These alternate at theta frequency with sources (⇓) and sinks (⇑) in the corresponding layers. CSD laminar profiles were compared between theta cycles prior (blue traces) versus after the stimulus (red traces). No significant difference was detected between the generating profiles of spontaneous prestimulus and event-related poststimulus theta rhythms.
F. MUA recorded simultaneously and with the same contacts as the CSD in panel a. Firing in supragranular, granular and infragranular layers increases during the middle layer sink (⇓), and decreases during the middle layer source (⇑). Thus, the middle layer sink in panel a probably represents an EPSC rather than a passive current return.
G. Averaged theta phase. Phase in the theta range from the Hilbert transform is averaged with respect to stimulus onset. Peaks at ~100, 300, 600 and 800ms (⇓) indicate phase-resetting of theta activity by word onset.
H. Average evoked PG response. Phase-resetting results in averaging being able to reveal an evoked response, despite the strong decrease in overall LFP amplitude.
I. Spectral power, averaged on the peak of the sink in Layer III, increases in a wide band up to ~150Hz. Scale same as C except pixels with zscore less than 3 have been masked as green.

A. Yellow arrow indicates the entorhinal cortex (ER) laminar probe location on a sagittal MRI taken with the probes in situ (Patient 3). Laminar tip is in the medial aspect of the left parahippocampal g., adjacent to the cisterna ambiens () (coordinates −23, −17, −25).
B. iERSP evoked in the WR task. Initially, words evoke strong wideband (~2-200Hz) increases in spectral power (↓). The increase is present in both superficial and deep layers, to both new and old stimuli, but is strongest in superficial layers to old words (↓↓). In most frequencies, the broadband increase appears to begin at ~300ms after word onset. Significant old> new differences are present in all frequency bands from ~400-600ms after word onset, mainly in superficial layers. This broadband increase is followed by a theta band increase, from ~500-1100ms in upper layers, new (↑) greater than old (↑↑). At still longer latencies, after 1100ms, theta is larger to old stimuli in both superficial (⇑) and deep layers (Δ). Although the first burst of high frequency activity is mainly to old words and over by ~600ms, high frequency activity continues, especially to new words, until ~2000ms (^).
C. Phase-locking between superficial and deep CSD. Phase locking increases mainly in gamma frequencies to new words (⇓), and mainly in theta and delta bands to old words (↓).
D. PG waveforms from individual trials of WR, aligned on stimulus onset and superimposed. Single trials show an initial component with duration of ~400ms peaking at ~300ms (⇓), followed by a slower components with peaks at ~750ms (⇑) and 1700ms (↓).
E. Inter-Trial Coherence of PG in putative deep ER layer. ITC in low theta and delta frequencies increases relative to baseline until ~900ms after stimulus onset (⇑).
F. Event-related averaged PG during WR. Phase-synchronization of low frequency PG waveforms allows them to contribute to the cross-trial average.
G. Depth profile of the theta CSD. Broadband CSD was averaged to theta peaks found in the band-passed data. Distinct profiles are observed for theta triggered to peaks in the superficial (red traces) versus deep (black traces) layers. In both cases, there are two spatially separated sinks, one deep (Δ) and one superficial (^), with each sink surrounded by its own current sources.

A. Approximate location of the laminar probe in the left superior prefrontal cortex (↓). Pt. 4.
B. Single sweeps of CSD from all channels, aligned on theta peaks. No differences are visible in the theta bursts from prestimulus (blue traces) versus poststimulus (red traces). Middle layer sinks (↑) and superficial sources (↓) are clearly visible in single sweeps.
C. CSD averaged with respect to theta peaks. The dominant CSD during theta is a middle layer sink (↑) paired with more superficial (↓) and deep (⇓) sources. These alternate with a middle layer source and superficial/deep sinks. No significant difference was detected between the generating profiles of spontaneous prestimulus (blue traces) and event-related poststimulus (red traces) theta rhythms.
D. MUA triggered simultaneously with the CSD in C shows increased neuronal firing across multiple cortical layers during the middle layer sink (↓).
E. Spectral power, averaged on the peak of the sink in Layer III, increases in a wide band up to ~150Hz.
F. iERSP evoked by new words in SJ. A delta-theta range increase begins at ~250ms and is most prominent in upper layers (⇓), but also occurs in middle (⇑) and deep layers (↓). In upper layers the theta continues to more than 900ms, and grows in the alpha range (⇓). In middle layers, the theta increase is followed by a decrease from ~600ms (↑). Unlike temporal lobe sites, no early broadband activation was evoked by words.
G. ITC shows increases in the low theta and delta frequencies relative to baseline from ~100-800ms after stimulus onset (⇑).
H. Event-related CSD average shows mainly delta frequencies.

A. Yellow indicates the tip of the right AC laminar probe in approximately Brodmann area 24b’ on a mid-sagittal MRI taken with the probes in situ (Pt 5). The third ventricle is dilated due to compensated aqueductal stenosis.
B. Laminar distribution of Theta. Broadband CSD averaged with respect to theta peaks. The dominant CSD during theta is a middle layer sink (↑) paired with more superficial source (↓). These alternate with a middle layer source (⇑) and superficial sinks (⇓). No significant difference was detected between the generating profiles of spontaneous prestimulus (blue traces) and event-related poststimulus (red traces) theta rhythms.
C. Inter-Trial Coherence shows increases in the delta frequencies relative to baseline from ~200-900ms after stimulus onset to old (⇓) and new words (⇑).
D. Event-related average shows mainly a delta frequency response.

CSD, averaged on the peak of the theta rhythm, shows that in all 4 cortical areas, the main currents generating the theta are a middle layer sink paired with a superficial source and a weaker deep source. Correlated increases in MUA (in two areas) imply that the sink is active excitation. Spontaneous (prestimulus) and evoked (poststimulus) theta distributions are identical, demonstrating that their neural generators are the same at a microphysiological level. Recordings from each cortical area was obtained in a different 4 subject.

A. Average sinks (red) and sources (blue) in different cortical layers are plotted against time during sentence reading. Orange and violet bars indicate the times of the individual words; Black vertical lines indicate word onsets. The initial word evokes a prominent layer IV sink (Δ, onset ~130ms, peak ~150ms) which overlaps with a slightly later layer III sink (⇓, peak ~160ms), followed by sinks in layers I and II (↓) from ~200-290ms. Responses to later words are similar except for a much attenuated initial layer IV sink. Sentences contained either 11 or 13 words. Rhythmic transmembrane currents continue after the sentence (⇑). Note that there is a consistent layer III/IV sink (↑) occurring to all words except the initial, at ~40ms, prior to the arrival of information from the current word.
B. Average CSD amplitude from one channel of the recording in panel A.
C. Average phase from 3 to 7 Hz of the CSD, showing phase-locking of theta to the word-presentation frequency.
D. High gamma power (HGP) in different cortical layers during sentence reading. The initial word evokes a sharp peak ([g314]) indicating increased firing at the same time as the layer IV sink (onset ~130ms, peak ~150ms). This sharp initial peak is also present in other layers, at slightly longer latencies, and is followed by a broad increase in firing from ~300-700ms (△). Sustained firing is also seen at the end of the sentence (⇑), confined to layer V.
E. Average CSD in different cortical layers evoked by Content versus Function words in sentences are compared. The initial layer II/IV sink (Δ) as well as the following I/II sink (↓) are larger to Content words.
F. Average CSD evoked by Function words that are commonly followed by Content versus Function words. Following the function words that predict content words, but prior to the arrival of neural information from the content word, there is a layer III/IV sink (↑) that is absent to words that follow function words which predict other functions words (^). This ‘predictive sink’ is visible (⇑, peak ~50ms) in the subtraction of the CSD evoked by the two conditions (FC-FF), as are the later evoked sinks in layers IV (⇓, peak ~150ms), and II (↓, peak ~220ms).

A. Sinks (red) and sources (blue) in different cortical layers are plotted against time for a single trial of the SJ task in Pt 1. Periods with superficial sinks and middle layer return sources (purple arrows) alternate with periods of superficial sources and middle layer sinks (orange arrows). Spontaneous oscillations preceding the stimulus are interrupted and reset by a sharp middle layer sink peaking ~200ms after word onset. After the stimulus, the laminar distribution of event-related sinks and sources resembles that of the spontaneous activity.
B. Waveforms from two of the channels (4 and 13) that were used to create the CSD color map in A illustrate the alternating sinks and sources in middle and superficial layers.
C. Schematic neurons showing the relationship hypothesized between CSD sinks and synaptic activity. The middle layer sink (left) is interpreted as excitatory feedforward synapses (EPSCs) from lower cortical areas, whereas the superficial sink (right) is interpreted as feedback synaptic activity from higher cortical areas as well as local associative input. CSD estimates transmembrane current, shown entering the cell at the sites of the EPSCs (white plus signs).
D. Consistent with the location of the EPSCs closer to the soma, the middle layer sink is associated with increased population neuronal firing. The spread of new information from lower cortical areas may be facilitated by relative excitatory modulation (left). Over time, as the network becomes increasingly dominated by local and distant feedback associations (right), relative inhibitory modulation requiring convergent activation would prevent run-away excitation.