{"id":"0a1006bb-611e-4909-9d5c-ca594803dd50","arxiv_id":"2607.08855","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"SNST recovers consistent amplitude-envelope and slow-rhythm-gated coupling in motor-imagery EEG that phase-lag indices do not detect under matched FDR control.","lead":"SNST extends wavelet scattering to multichannel EEG so it measures how strongly neighboring channels co-modulate in amplitude, and how that coupling is gated by slower rhythms. On a standard motor-imagery dataset it finds a consistent centro-parietal pattern that phase-lag measures miss under the same statistics, giving a complementary connectivity signal for BCI and clinical EEG.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.5","headline":"The zero-overlap dissociation with PLI/wPLI is load-bearing for the 'distinct connectivity signal' claim, but residual volume conduction after Laplacian can still produce amplitude co-modulation that phase-lag measures are designed to ignore.","rationale":"The Reader correctly isolates residual volume conduction / shared noise after Laplacian as the weakest assumption supporting the central claim. My stress-test simply sharpens the same point: the zero-overlap result with PLI/wPLI is not independent evidence of a distinct physiological signal; it is the expected signature of any residual common-source amplitude co-modulation that phase-lag indices are designed to suppress. Spatial consistency across subjects does not adjudicate the alternative. The concrete synthetic common-source test after the exact Laplacian would settle whether residual mixing alone can reproduce the reported pattern. Because the paper already flags residual correlation and limits itself to a single 22-channel dataset, the Reader’s CONDITIONAL verdict (accept-shaped once residual-VC and generalizability are addressed) remains the right call; no upgrade or downgrade is warranted. The second-order floor and small n are secondary relative to this load-bearing ambiguity.","tokens_in":14650,"tokens_out":742,"duration_ms":8175,"concrete_test":"Generate synthetic multi-channel epochs that share a single cortical source mixed through a realistic lead-field (or spherical volume-conduction model) at the exact 22-electrode montage, apply the identical spherical-spline Laplacian (m=4, λ=10^{-5}), then compute S_norm1, PLI and wPLI under the paper’s bias-correction + FDR pipeline. If a non-negligible fraction of the real-data significant S_norm1 pairs (especially C4–CP4 and immediate neighbors) still survive FDR while PLI/wPLI remain near zero, residual volume conduction is sufficient to explain the dissociation and the 'distinct signal' claim weakens.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The strongest claim rests on two linked results: (i) S_norm1 finds spatially consistent central-parietal coupling across all 9 subjects and both conditions, and (ii) under the identical bias-corrected FDR pipeline, PLI and wPLI recover essentially nothing with zero overlap, so amplitude-envelope coupling is 'largely distinct.' The paper correctly notes that residual inter-channel correlation remains after Surface Laplacian (§3.1: post-Laplacian mean pairwise r still 0.249–0.296; §4: residual spatial dependence may inflate independent findings). Because S_norm1 is the time-averaged product of same-scale analytic envelopes (Eqs. 2–4), any residual common-source or shared-noise component that survives the Laplacian will raise envelope co-fluctuation without producing a consistent non-zero phase lag. PLI/wPLI are constructed to discard exactly that component. Therefore the observed zero-overlap dissociation is expected even if a non-trivial fraction of the S_norm1 signal is residual volume conduction rather than genuine inter-regional amplitude coupling. Spatial consistency across subjects does not break this alternative: the same residual mixing geometry is present for every subject. The synthetic-noise checks described in §2.4 only plant independent noise or pure planted coupling; they do not quantify residual common-source leakage after the exact Laplacian used on the real data. Without that control, the claim that SNST recovers a connectivity domain 'structurally inaccessible' to phase measures remains under-supported.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.5","summary":"The manuscript introduces the Spatial Neighboring Scattering Transform (SNST), a multichannel extension of the wavelet scattering transform that replaces the single-channel modulus with a cross-channel conjugate product. This yields two descriptors: S_norm1, quantifying same-scale amplitude-envelope co-modulation between electrode pairs, and S_norm2, quantifying slower-scale modulation of that coupling. On BCI Competition IV-2a (n=9), after Surface Laplacian preprocessing and a bias-corrected one-sample t-test with Benjamini–Hochberg FDR, S_norm1 identifies spatially consistent central-parietal coupling (notably C4–CP4 and neighbors) across all subjects and both left- and right-hand imagery; S_norm2 indicates gating by slow rhythms below ~11 Hz. Under an identical correction pipeline, PLI and wPLI recover at most three significant combinations with zero overlap with SNST, which the authors interpret as evidence that amplitude-envelope coupling is a largely distinct connectivity signal inaccessible to phase-lag measures.","tokens_in":15107,"tokens_out":1624,"duration_ms":23863,"significance":"If the residual-artifact alternative can be ruled out and the method is shown to add information beyond standard amplitude-envelope correlation, SNST would be a useful, deformation-stable descriptor for amplitude-domain and cross-frequency EEG connectivity, with clear equations, synthetic null/planted checks, and matched FDR comparison to PLI/wPLI. The second-order descriptor (modulation of inter-channel coupling strength) is conceptually distinct from classical phase–amplitude CFC and is the more novel contribution. Explicit formulae (Eqs. 2–6), volume-conduction checks for PLV/PLI/wPLI, and an honest limitations section are strengths. The work is currently a single-dataset, sensor-level proof of concept rather than a fully established connectivity standard.","major_comments":[{"comment":"§3.1–§3.4 and §4: The load-bearing claim that SNST recovers a ‘largely distinct’ genuine inter-regional connectivity signal rests on zero overlap with PLI/wPLI under matched FDR. S_norm1 (Eqs. 2–4) is the time-averaged product of same-scale analytic envelopes; residual common-source or shared-noise components that survive the Laplacian will raise envelope co-fluctuation without producing consistent non-zero phase lag. Post-Laplacian mean pairwise correlation remains 0.249–0.296 (§3.1), and the paper itself notes residual spatial dependence may inflate independent findings (§4). PLI/wPLI are constructed to discard exactly that component, so zero overlap is expected even if a non-trivial fraction of S_norm1 is residual mixing. Spatial consistency across subjects does not break this alternative (same montage/mixing geometry for all subjects). The synthetic checks in §2.4 use independent noi","section":null},{"comment":"§2.5 and §3.4 / Table 1: The only external benchmarks are phase-based (PLV, PLI, wPLI). For a paper whose central product is amplitude-envelope coupling, the natural comparators are established amplitude measures—amplitude envelope correlation (AEC) and orthogonalized AEC—not only phase-lag indices. S_norm1 is closely related to scale-wise envelope co-fluctuation; without a head-to-head comparison (same epochs, same FDR pipeline, same seeds), it is unclear whether SNST recovers structure beyond standard AEC or mainly re-expresses it in a scattering framework. The novelty claim for first-order SNST and the interpretation of zero phase-overlap as a new connectivity domain both depend on this missing comparison.","section":null},{"comment":"§2.3 Eqs. (5)–(6) and §3.3: S_norm2’s physiological interpretation as ‘slow-rhythm gating of cross-channel coupling’ is interesting but rests on an ad-hoc normalization (division by max(0.1·median S1, ε)) and a paper-defined ‘breadth-of-modulation’ criterion. The Discussion links the <11 Hz range to slow cortical potentials / Bereitschaftspotential without simultaneous SCP recording. Either strengthen the claim with a direct comparison to a standard cross-frequency amplitude–amplitude coupling metric on the same data, or soften the gating language to a descriptive report of second-order coefficient structure and treat the SCP link as a hypothesis for future work.","section":null},{"comment":"Validation design (§2.4, Results §3.2): Primary validation is spatial consistency of FDR-significant pairs across n=9 subjects on a single 22-channel public dataset. That is a reasonable exploratory criterion but is soft relative to the abstract/conclusion language that SNST ‘establishes’ a systematically recoverable connectivity domain ‘applicable to any multichannel EEG analysis.’ At minimum, the claims should be scoped to this dataset and montage, and either a second dataset or a clearer leave-one-subject-out / split-session consistency analysis should be added so that the central-parietal pattern is not solely group-averaged on the same nine subjects used for discovery.","section":null}],"minor_comments":[{"comment":"Figure 6 caption begins with ‘eed-based’ (missing ‘S’); fix typography.","section":null},{"comment":"Figure 2 caption: ‘same timed epoch’ → ‘same timed epoch’ / ‘same time epoch’ wording is unclear; clarify that pairs are formed within the same epoch across channels.","section":null},{"comment":"§2.3: State explicitly that |(z_m)(z_n)*| = |z_m||z_n| so readers immediately see U1 as an envelope product, not a phase-sensitive cross-spectrum term.","section":null},{"comment":"Free parameters (J1,Q1,J2,Q2, Laplacian m/λ, S_norm2 floor 0.1) are listed but sensitivity is not reported; a brief supplementary sweep or justification that results are stable under modest changes would help.","section":null},{"comment":"Abstract and Conclusion slightly over-claim relative to the single-dataset offline analysis; align wording with the limitations already stated in §4 (online classification and multi-dataset generalization left for future work).","section":null},{"comment":"Reference list and in-text citations are generally appropriate; consider citing orthogonalized AEC / Hipp-style amplitude-correlation connectivity more prominently when framing the amplitude-domain gap.","section":null}],"recommendation":"major_revision","confidential_remarks":"The residual-volume-conduction critique is the main reason I recommend major revision rather than minor: the PLI/wPLI dissociation is partly by construction and does not by itself establish genuine inter-regional amplitude coupling. The missing AEC comparison is equally important for novelty assessment. If the authors add residual-mixing controls and AEC benchmarks and scope the claims, this could become a solid methods contribution; without them the central ‘distinct connectivity signal’ claim remains under-supported for a serious methods/neuroscience journal. Scope fit is fine for q-bio.NC / EEG methods venues; impact is currently moderate."},"author_rebuttal":null,"desk_editor":{"model":"grok-4.5","letter":"The one thing worth knowing: they replace the single-channel modulus in Mallat/Andén scattering with a cross-channel conjugate product and get two usable descriptors (Snorm1 for envelope co-modulation, Snorm2 for its slow-frequency gating). That is a genuine technical extension, not present in the single-channel EEG scattering papers they cite, and they write the equations cleanly (Eqs. 2–6).\n\nWhat they do well: matched bias-correction and BH-FDR against PLI/wPLI on the same epochs, synthetic null and planted-coupling checks, volume-conduction stress test of the phase measures, and an honest limitations section. On BCI IV-2a (n=9) they find a centro-parietal neighborhood (C4–CP4 and neighbors) that is significant and spatially consistent across all subjects and both hands, with essentially zero overlap to PLI/wPLI. That dissociation is real under their pipeline and supports the claim that amplitude-envelope coupling is a largely distinct connectivity signal from phase-lag measures.\n\nSoft spots, in proportion. The stress-test concern lands: post-Laplacian mean pairwise r is still 0.25–0.30, and Snorm1 is exactly the quantity residual common-source leakage will inflate without producing a consistent phase lag. Spatial consistency across subjects does not break that alternative—the mixing geometry is shared. Their synthetic checks do not quantify residual leakage after the exact Laplacian used on the real data, so the stronger language (“structurally inaccessible,” “genuine inter-regional”) is under-supported. Second-order floor is ad-hoc; single public 22-channel dataset; no code. None of that kills the methods contribution, but it caps the physiological claim.\n\nThis is for people who already care about amplitude-domain EEG connectivity or BCI feature design. Math and citation pattern look solid; circularity is low. I would send it to peer review—methods papers that ship explicit operators and matched controls deserve referee time even when the residual-volume-conduction control is incomplete. Engage if you work on envelope coupling or scattering features; wait for multi-dataset or residual-mixing controls before treating the “distinct connectivity domain” claim as settled.","headline":"Clean multichannel scattering extension that recovers amplitude-envelope coupling with a real PLI/wPLI dissociation; residual volume conduction after Laplacian still undercuts how strongly we can call that signal \"genuine inter-regional.\"","tokens_in":15749,"tokens_out":553,"would_cite":false,"duration_ms":5720,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.5","headline":"A new EEG connectivity measure finds amplitude-envelope coupling that phase methods miss, consistently in motor imagery.","keywords":["Spatial Neighboring Scattering Transform","wavelet scattering transform","amplitude-envelope coupling","EEG functional connectivity","cross-frequency coupling","motor imagery","phase lag index"],"falsifier":"Recompute SNST first-order coefficients on the same epochs after source reconstruction or denser high-density EEG with stricter spatial filtering; if the central-parietal significant neighborhood disappears or collapses into volume-conduction patterns also recovered by PLI/wPLI, the claim of a distinct amplitude-domain connectivity signal fails.","tokens_in":15535,"feed_emoji":"🧠","tokens_out":662,"duration_ms":5737,"temperature":0.7,"pith_summary":"Phase-based EEG connectivity measures discard amplitude information by design and can miss genuine inter-regional dependence expressed only through co-modulating envelopes. This paper introduces the Spatial Neighboring Scattering Transform (SNST), which extends the wavelet scattering transform to pairs of channels so that first-order coefficients capture amplitude-envelope coupling and second-order coefficients capture how that coupling is itself modulated by slower rhythms. On a public motor-imagery dataset, after Surface Laplacian filtering and a bias-corrected FDR pipeline, the first-order descriptor recovers statistically significant coupling concentrated in a central-parietal electrode neighborhood that is reproduced across all subjects and both left- and right-hand conditions. The second-order descriptor shows this coupling is gated by slow rhythms below about 11 Hz. Under the same statistical procedure, phase lag index and weighted phase lag index recover almost no significant pairs and share zero overlap with SNST, supporting the claim that amplitude-envelope coupling is a largely distinct connectivity signal. A sympathetic reader cares because many clinical and BCI analyses currently ignore this signal by construction.","feed_headline":"EEG measure finds amplitude coupling phase methods miss","feed_subtitle":"Scattering transform on channel pairs recovers central-parietal coupling gated by slow rhythms in motor imagery","key_machinery":"Spatial Neighboring Scattering Transform (SNST): replace the single-channel modulus of the wavelet scattering transform with a cross-channel conjugate product of band-pass filtered signals, then low-pass average and normalize, yielding first-order envelope-coupling coefficients and second-order coefficients that describe slow modulation of that coupling.","core_discovery":"SNST's first-order descriptor identifies statistically significant amplitude-envelope coupling within a central-parietal electrode neighborhood that is spatially consistent across all subjects and both motor-imagery conditions; under an identical bias-corrected FDR pipeline, PLI and wPLI recover negligible significant coupling with zero overlap, so amplitude-envelope coupling constitutes a largely distinct connectivity signal from phase synchronization.","pith_inferences":[],"forward_implications":[],"fun_headline_variants":["SNST recovers amplitude coupling PLI/wPLI miss in motor imagery EEG","Central-parietal amplitude coupling gated by slow rhythms via SNST","Scattering transform finds amplitude-envelope links phase measures ignore","SNST shows cross-subject central-parietal coupling distinct from phase sync","First-order SNST maps consistent amplitude coupling across all MI subjects"],"cache_read_input_tokens":128,"weakest_assumption_plain":"That spatial consistency of FDR-significant electrode pairs across subjects, after Surface Laplacian filtering, is enough to treat the coupling as genuine inter-regional amplitude dependence rather than residual volume conduction or shared noise on a single 22-channel dataset.","fun_headline_variants_meta":{"raw":{"variants":["SNST recovers amplitude coupling PLI/wPLI miss in motor imagery EEG","Central-parietal amplitude coupling gated by slow rhythms via SNST","Scattering transform finds amplitude-envelope links phase measures ignore","SNST shows cross-subject central-parietal coupling distinct from phase sync","First-order SNST maps consistent amplitude coupling across all MI subjects"]},"model":"grok-4.5","effort":"low","cost_usd":0.001898,"raw_usage":{"total_tokens":891,"prompt_tokens":814,"num_sources_used":0,"completion_tokens":77,"cost_in_usd_ticks":18980000,"prompt_tokens_details":{"text_tokens":814,"audio_tokens":0,"image_tokens":0,"cached_tokens":128},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":0,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":814,"tokens_out":77,"duration_ms":3069,"temperature":1.0,"reasoning_tokens":0,"cache_read_input_tokens":128,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-13T06:14:22.471716+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"Recompute SNST first-order coefficients on the same epochs after source reconstruction or denser high-density EEG with stricter spatial filtering; if the central-parietal significant neighborhood disappears or collapses into volume-conduction patterns also recovered by PLI/wPLI, the claim of a distinct amplitude-domain connectivity signal fails.","supporting_citations":[],"review_version":1}