{"id":"009cbc64-b189-4602-9215-448514fda798","arxiv_id":"2608.10560","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Spectral distinguishability bounds the number of frequency labels a biological oscillator can carry by its quality factor Q = 2 pi nu tau, and applying this screen eliminates high-frequency molecular carriers while leaving low-frequency neural rhythms.","lead":"A new screening framework argues that any biological oscillator used as a frequency label is limited to at most Q = 2 pi nu tau distinguishable values, so only low-frequency brain rhythms can plausibly serve as labels. The framework offers a cheap, substrate-independent way to evaluate exotic proposals such as microwave or terahertz information carriers in the brain.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Eq. (1)'s central bound is contradicted by the paper's own C2 and Table 2: with SNR above 6, Cramér-Rao estimation resolves more than Q labels, so Q is not an information-theoretic limit.","rationale":"The reader flagged the B ∼ ν assumption in Eq. (1) as the weakest point. That is a real overstatement in the abstract, but it is not the most load-bearing issue: for a positive-frequency carrier the usable band is at most 2ν, so B ∼ ν is only off by a factor of two, and it is explicitly stated rather than hidden. The more serious problem is internal. Section 1.2 says two instances are distinguishable 'only if' their frequencies differ by more than the linewidth; Section 3.1 then derives a Cramér-Rao estimation floor that is smaller than the linewidth whenever the SNR exceeds 6, and Table 2 exploits exactly this to resolve M = 100 labels on a carrier with Q ≈ 38. These two statements cannot both be true as a universal bound. The distinction between 'source' and 'reader' does not save Eq. (1), because distinguishability is a property of the readout: if a reader can resolve 100 frequencies, the carrier supports 100 labels. The paper's own Table 2 is the cleanest witness. The 30 GHz rejection in §2 remains robust, since Q = 0.19 < 1 means the mode is overdamped and no readout scheme can recover a reproducible frequency; the geometric no-cavity argument and the power budget are independent of this flaw. The screen's C5 also carries much of the weight in excluding high-frequency carriers. So the paper is not worthless, but its central theoretical claim needs to be reframed as a heuristic under a Rayleigh-resolution/low-SNR criterion, not an information-theoretic bound. That is a significant revision, consistent with the reader's CONDITIONAL verdict; hence the final recommendation is unchanged from the reader's.","tokens_in":14215,"tokens_out":14142,"duration_ms":148694,"concrete_test":"Recompute Table 2 adding a column showing the maximum M allowed by Eq. (1) for the 40 Hz, 0.15 s carrier (Q ≈ 37.7). Row M = 100 has required ρ = 42; Eq. (4) then gives σ_f = 0.40 Hz = ν/M, so the row claims 100 resolvable labels despite Q ≈ 38. If the reproduced table confirms this, C1 and C2 are mutually inconsistent and Eq. (1) is not a hard bound.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Eq. (1) is presented as a universal bound from 'spectral distinguishability alone', but §3.1 (C2) contradicts it. Eq. (1) assumes two frequencies are resolvable only if separated by the linewidth δν = 1/(2πτ_coh), giving M ≤ Q. Eq. (4) (Cramér-Rao) gives σ_f ≥ 1/(2πT)√(6/ρ); at T = τ_coh this is σ_f = δν√(6/ρ). For ρ > 6 the estimation error is below the linewidth, so a reader can resolve more than Q frequency values from the same source. The paper's own Table 2 is a counterexample: for a 40 Hz carrier with τ_coh = 0.15 s (Q ≈ 38), it lists M = 100 as resolvable at ρ = 42, with σ_f = 0.40 Hz equal to the channel spacing ν/M. Thus its own criterion allows 100 > 38 labels. Saying C2 'is not C1 restated' does not resolve this: C1's 'only if' is itself a distinguishability claim, and C2 is the quantitative statement of distinguishability. Eq. (1) is therefore at best a low-SNR, Rayleigh-resolution heuristic, not a bound on how many labels a carrier can support; the abstract's 'spectral distinguishability alone bounds' overstates it. The §2 rejection of the 30 GHz proposal survives because Q < 1 there is an overdamped regime, but any use of Q as an information-theoretic capacity must be revised.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes that any frequency-multiplexed biological oscillator can carry at most M ≤ Q = 2π ν τ_coh distinguishable labels, deriving this from the relation between linewidth and coherence time together with the assumption that the usable frequency band scales with the carrier frequency (B ~ ν). It applies this bound to reject a recently proposed 30 GHz intracolumnar microwave field, derives six further criteria (readout, coupling, power, persistence, writability, thermal regime), screens eleven candidate carriers in Table 3, and concludes that only low-frequency neural rhythms pass. The paper includes a detailed reproduction of the source model's arithmetic, an openly archived reference implementation, and explicit statements of the framework's limitations.","tokens_in":14514,"tokens_out":7017,"duration_ms":72113,"significance":"If the central bound were correct as stated, it would provide a substrate-independent, two-parameter screen for proposed biological information carriers and would explain why biology uses frequency coding only for small alphabets. The manuscript has notable strengths: Appendix A carefully reproduces the Keppler model's numbers, the scripts that regenerate all tables are openly archived, the falsification conditions in §6.4 are concrete, and §6.3 explicitly flags the circularity in the Q-versus-ν conjecture. However, the headline claim is compromised by an internal contradiction with the paper's own C2 criterion and Table 2, and by the unproven B ~ ν assumption. The qualitative screening conclusions may survive a revision, but the abstract's statement that 'spectral distinguishability alone bounds the number of labels' is not supported by the derivation as it stands.","major_comments":[{"comment":"The central bound M ≤ Q is contradicted by the paper's own readout criterion. Eq. (4) gives σ_f ≥ (1/2πT)√(6/ρ); at T = τ_coh this is σ_f = δν√(6/ρ), so for ρ > 6 the estimation error is smaller than the linewidth. Table 2 makes the contradiction concrete: for a 40 Hz carrier with τ_coh = 0.15 s, Q = 38, yet the table lists M = 100 as resolvable at ρ = 42, with σ_f = 0.40 Hz equal to the channel spacing ν/M. Thus the paper's own criterion allows more than Q labels from the same source. Saying that C2 'is not C1 restated' does not resolve this, because Eq. (1) is justified by the statement that two instances are distinguishable only if their frequencies differ by more than the linewidth; C2 is the quantitative statement of distinguishability under readout. At best, Eq. (1) is a Rayleigh-resolution heuristic for simultaneous tones at low SNR, not a bound on the number of labels a single carrier can support. The abstract and §1.2 should be revised to state the conditions (e.g., no readout averaging, B ~ ν) under which Q bounds M, or the framework should be reformulated around a capacity expression that includes SNR and observation time.","section":"§1.2, Eq. (1); §3.1, Eq. (4); Table 2"},{"comment":"The bound depends critically on the assumption B ~ ν, which is asserted but not derived. The text calls this 'the most favourable assumption available to any such proposal,' but no argument is given that a biological oscillator could not use a band wider than its center frequency. For example, a relaxation oscillator or a chemically tunable oscillator could conceivably sweep over a range broader than ν, in which case the number of resolvable frequency values would not be capped by 2πντ_coh. Because the abstract claims that 'spectral distinguishability alone' produces the bound, the role of the B ~ ν assumption is understated. The paper should either prove or explicitly conditionalize the bound on B ~ ν and justify why this is the appropriate bandwidth for the carriers under discussion.","section":"§1.2"},{"comment":"The text states that 'C2 is therefore usually not binding when C1 is satisfied' and that 'improving signal-to-noise buys nothing until coherence time improves.' This is contradicted by Table 2, where, at fixed τ_coh = 0.15 s, increasing ρ from 0.42 to 42 increases the resolvable channel count M from 10 to 100. The sentence appears to mean that C2 does not impose a lower bound beyond C1 in the regime where Q is already satisfied, but as written it is false and it also undercuts the claim that a proposal failing C1 cannot be rescued by a sensitive reader. For carriers with Q > 1, a high-SNR reader can resolve more than Q labels, so the paper's own criterion provides a rescue route that should be explicitly addressed.","section":"§3.1; Table 2"}],"minor_comments":[{"comment":"In the 1 ms row, the linewidth is listed as 1591 Hz, but 1/(2π × 0.001 s) ≈ 159 Hz; the reported Q = 1.9 × 10^8 is consistent with 159 Hz, so this appears to be a typographical error that should be corrected.","section":"Table 1"},{"comment":"The coherence times are presented without error bars or confidence intervals, and §7 acknowledges that they are often estimated rather than measured. Adding quantitative uncertainty ranges to the table would make the screen more informative and would help readers assess the robustness of the pass/fail classifications.","section":"Table 3"},{"comment":"There are several formatting artifacts, such as 'T able' instead of 'Table' (Tables 1, 2, 3, A.1) and 'F requency' instead of 'Frequency' (Table 4). These should be cleaned up in a final revision.","section":"Throughout"},{"comment":"The paper is commendably explicit that the Q-versus-ν conjecture is partly circular and that Eq. (8) is insecure. However, because Eq. (8) appears in the 'design rule' subsection and is used to motivate an account of gamma as the highest usable frequency, it would be helpful to mark it more visibly as a non-result in the main text, perhaps by moving it to a clearly labeled speculative paragraph.","section":"§6.3"}],"recommendation":"major_revision","confidential_remarks":"The contradiction between Eq. (1) and the paper's own Table 2 is the main obstacle: the central bound is stated as universal but is actually a low-SNR, proportional-bandwidth heuristic. This is fixable by reframing C1 as a source-side Rayleigh bound and by adding explicit conditions (bandwidth scaling, SNR, observation time) to the theorem. The case study of the 30 GHz proposal and the multi-criteria screen are valuable and likely to survive such a revision. I recommend major revision rather than rejection, and I encourage the editor to send the revised version back to a referee familiar with estimation theory to verify the revised capacity statement."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read this one. The seven-criterion screen is genuinely useful, and the worked rejection of the 30 GHz microwave proposal is the strongest part. But the headline bound M≤Q is not the universal limit the abstract claims; the paper's own C2 criterion contradicts it.\n\nThe good: the persistence window (C5) and writability (C6) are real contributions. C5 cuts both ways—coherence too short is useless, but coherence too long prevents rewiring. C6 forces proposals to say how the label is set, which is usually missing. The microwave rejection uses geometry and power budgets, not decoherence assumptions, so it is robust. The paper is self-aware: it flags Eq. (8) as insecure and admits the Q-ν conjecture is partly circular. The archived reference implementation regenerates every table.\n\nThe problem: Eq. (1) states two frequencies are distinguishable only if separated by the linewidth. But C2 (Cramér-Rao) shows that with SNR above 6, a reader can estimate a frequency more accurately than the linewidth. Table 2 itself lists M=100 for a 40 Hz carrier with Q≈38 at SNR 42. So the paper's own numbers allow more than Q labels. Q is a Rayleigh-resolution heuristic, not an information-theoretic bound. The 'spectral distinguishability alone' claim in the abstract overstates it. The B~nu scaling is also assumed, not derived, though the paper does say it is the 'most favourable assumption'.\n\nThe 30 GHz rejection survives because Q<1 there means the mode does not complete a cycle—that is a physical statement independent of SNR. And the screen's conclusion that high-frequency carriers fail is likely robust, because C5 kills them anyway. But the general bound needs revision.\n\nFor a reader, this is worth a serious referee: the framework is useful and the thinking is honest, but the central theorem is mis-sold. A revision that reframes Eq. (1) as a low-SNR heuristic and reconciles C1 with C2 would make it solid. I would send it out, but the author should not publish the bound as a hard limit.","headline":"Useful screening framework, but the headline bound M≤Q is overclaimed and contradicted by the paper's own SNR analysis.","tokens_in":15043,"tokens_out":5019,"would_cite":false,"duration_ms":52509,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":null,"created_at":"2026-08-12T21:48:36.219692+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":null,"supporting_citations":[],"review_version":1}