Contribution
The first test of whether phase-amplitude coupling (PAC) in human fronto-parietal cortex tracks the allocation of visuospatial attention (p. 2: “no study to date has investigated if PAC may serve as a mechanism for visuospatial attention control within human fronto-parietal cortex”). With electrocorticography it shows that broadband high-gamma amplitude locks to the trough of an endogenous 2–5 Hz rhythm while attention is held, more strongly for the contralateral field, and that trial-by-trial coupling strength predicts reaction time.
Key insight
The slow rhythm is not entrained by the task. Targets were jittered (1000–2000 ms), so “the slower rhythms that we report here appear to be endogenous and are most likely determined by the local neurophysiology” (p. 10). What co-varies with behaviour is the coupling, not either band’s amplitude: reaction times “did not significantly correlate with mean amplitude of the delta/theta signal … or with mean amplitude of the HG signal” (pp. 4–5).
Assumptions
- High gamma indexes local spiking, and low-frequency phase indexes excitability (p. 2, citing refs 11, 12). The interpretation of PAC as “low-frequency phase controls neuronal excitability” is a proposed mechanism the paper adopts, not one it tests (p. 2).
- Patients. n = 8 (5 left, 3 right hemisphere), with epilepsy, off anti-epileptic medication for 2–3 days; channels with epileptiform activity or over later-resected tissue were removed (Table 1, pp. 10–11).
- PAC measure. Phase-locking value between the 2–5 Hz phase and the phase of the 100–150 Hz amplitude envelope refiltered at 2–5 Hz (Voytek et al.’s procedure, ref. 68), tested against 1000 time-shifted surrogates, FDR q < 0.01 (pp. 11–12). No control for the waveform shape of the slow rhythm is reported.
- Task. The Starry Night Test: covert attention to one hemifield, targets 62% on the cued side, dynamic distractors; 156 attended and 96 unattended trials per subject (p. 11).
Key results
- Behaviour (p. 2). d′ = 5.03 ± 0.68; hits 97.21%; mean RT 601.31 ± 28.64 ms; RTs slower with eccentricity, F(1,7) = 13.22, p < 0.01.
- High gamma tracks attention (p. 3). 263 of 875 electrodes (30.06%) had sustained HG increases; 47 of 263 were larger for the contralateral field, 3 of 263 for the ipsilateral. Alpha/beta (10–30 Hz) fell in 119 electrodes, all parietal or visual.
- PAC (p. 3). 123 of 273 electrodes (45.05%, all eight subjects) showed significant 2–5 Hz phase to 80–250 Hz amplitude coupling during the 0–1000 ms allocation period, PLVs 0.16–0.57. (The paper gives the HG electrode count as 263 on p. 3 and as the 273 PAC was tested on, also on p. 3; the discrepancy is not explained.)
- Laterality (p. 4). 55 of 123 had higher PAC for contralateral attention; none higher for ipsilateral. Only 2 of the 55 showed a matching delta/theta power difference, so “power differences between conditions cannot account for the PAC differences”.
- Not an evoked-potential artefact (pp. 4, 7–10). Using only the last 1000 ms of long trials, 107 of 273 electrodes stayed significant, all among the original 123; HG locked to the trough in 117 of 123 for at least three cycles; ERPs and PAC overlapped in only 34 electrodes.
- PAC predicts RT (p. 4, Fig. 4). In 28 of 123 electrodes, in seven subjects (frontal 42.86%, parietal 32.14%, visual 25.00%), stronger PAC in the 1000 ms before the target went with faster RTs: S3 r(41) = −0.51, p < 0.01; S5 r(47) = −0.37, p < 0.05 (FDR-corrected), for contralateral attention only; ipsilateral r(40) = −0.11 and r(47) = −0.10, n.s.
- Microsaccades are not ruled out (p. 10); argued unlikely because the effects are lateralized and above 70 Hz.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | High-gamma amplitude couples to 2–5 Hz phase in human fronto-parietal and visual cortex during held spatial attention | strong | 123/273 electrodes in 8 subjects, surrogate-tested, robust to removing the ERP window (pp. 3–4) |
| C2 | The coupling is stronger for the contralateral field | moderate | 55/123 electrodes, none reversed; power controlled (p. 4) |
| C3 | Trial-by-trial coupling strength predicts reaction time | moderate | 28/123 electrodes; within-subject correlations, r ≈ −0.37 to −0.51 in the shown cases (p. 4) |
| C4 | PAC “serves as a mechanism to facilitate processing” in attention | weak | interpretation of C1–C3 (p. 9); no manipulation of PAC |
| C5 | PAC indexes “the overall level of engagement” of fronto-parietal cortex | weak | the closing reading of C3 (p. 10) |
Concepts
- Phase-amplitude coupling (PAC) — “the statistical dependence between the phase of a low-frequency rhythm and the amplitude (or power) of the high-frequency component” (p. 2). Defined as a dependence, not as control.
- High gamma (HG) — broadband 70–250 Hz power, taken as a correlate of local spiking (p. 2).
- Trough-locking — HG power peaking at the trough of the slow signal (p. 4).
Connections
- Canolty and colleagues (ref. 19) first showed theta–HG coupling in humans; Lakatos and colleagues (ref. 43) showed delta-phase modulation in macaque V1 under rhythmic attention; the present study differs by jittered targets and association cortex (pp. 9–10).
- The PAC method is Voytek and colleagues’ (ref. 68); Voytek is a co-author of Donoghue et al. 2021 (LIT-1999, NOTE-1529), which warns that waveform shape can produce PAC.
- Rodent theta–gamma coupling tracking learning (ref. 63) is cited as further evidence that “PAC serves a functional role” (p. 10).
Bearing on the record
- Jennings 2026 (LIT-1963, NOTE-1508) overstates it in one word. Jennings writes that PAC “has been shown to be associated with and to improve attention (Szczepanski et al., 2014)” (manuscript p. 28). The paper shows association: coupling strength correlates with RT in 28 electrodes. It does not manipulate coupling, so it cannot show that coupling improves attention; “facilitate” in its title is the authors’ interpretation (C4). “Associated with” is faithful; “improve” is not.
- Nor does it show that the slow wave constrains the fast one. Jennings calls PAC “evidence that low frequency waves constrain high frequency waves, a form of power or control” (p. 28). The paper defines PAC as a statistical dependence (p. 2) and takes excitability control from earlier work as the “physiologically plausible mechanism”, a mechanism wholly at the level of membrane potentials and spiking. Nothing in it favours the contextually emergent reading of THEORY-1314 over a micro-level one; it bears out that entry’s note that the signature fits weak emergence too.
- THEORY-1341. The task is covert, cue-directed spatial attention, i.e. selection by an instructed goal. It says nothing about fluid interests, stalemates or dlPFC feedback as a criterion; its frontal electrodes are not localized to dlPFC as such.
- The exchange’s (A5) methodological caution is apt but untested here. The paper reports no check of the slow rhythm’s waveform shape (NOTE-1529). Its coupled band, 80–250 Hz against a 2–5 Hz phase, is broad and far above low harmonics, which by Donoghue et al.’s own heuristic points toward genuine coupling rather than a shape artefact, but that is an inference, not a control.
- No instruction for machine-learning practice.
Limitations
- Eight patients with epilepsy, unequal hemisphere coverage; grids at 1 cm spacing.
- Correlational throughout; no stimulation or perturbation of the rhythm.
- The behaviour correlation holds in 28 of 123 coupled electrodes; the paper reports two examples, not the distribution of r.
- No control for waveform shape or aperiodic activity; microsaccades not excluded (p. 10).
- The 263/273 electrode count discrepancy on p. 3.
Open questions
- Does entraining or disrupting the 2–5 Hz rhythm (e.g. by rhythmic stimulation) change reaction times, as “facilitate” implies?
- Does the coupling survive cycle-by-cycle waveform-shape controls?