{"id":"cab315cd-1726-4a85-a509-53e2d22ec5d9","arxiv_id":"2507.13570","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"Coupling strength and temperature drive coupled denatured Morris-Lecar neurons through chaos, quasi-periodicity, synchronized bursting, and decay oscillations.","lead":"This paper simulates small networks of a simplified neuron model and tracks how their behavior changes with coupling strength and temperature. It reports chaos, quasi-periodicity, synchronized bursting, and decay oscillations, and offers a reproducibility kit.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The quasi-periodic regime at weak coupling is inferred only from intermediate 0-1 test values and visual inspection; no direct quasi-periodicity diagnostic (Lyapunov spectrum, frequency ratio) is computed.","rationale":"The reader's weakest assumption concerns the reliability of the 0-1 test and Hurst exponent, especially the discarding of negative K and H values. That concern is valid and adjacent to mine, but it does not capture the most specific gap: the positive claim of quasi-periodicity at weak coupling rests on intermediate K values and visual inspection, with no direct test for a second incommensurate frequency or zero Lyapunov exponents. If the Lyapunov spectrum at θ=-1 (gap junction) or θ=-0.5/-0.1 (Josephson junction) shows a positive exponent, the abstract's route from chaos to quasi-periodicity to synchronized bursting would be factually wrong at the quasi-periodic step; if two zero exponents are found, the concern is resolved and the central qualitative finding stands. The paper has reproducible code and an honest presentation of spurious metric values, which supports a conditional rather than a rejection verdict. Since the reader already recommended CONDITIONAL, my analysis does not change the verdict, but it sharpens the specific condition that should be met before the central claim is treated as established.","tokens_in":40671,"tokens_out":4541,"duration_ms":56491,"concrete_test":"For θ=-1 of model (3.1) and for θ=-0.5 and -0.1 of model (3.5), compute the full Lyapunov spectrum using Benettin's method over t∈[0,40000] with reorthonormalization every few integration steps, using the same parameters and initial conditions as Section 5. For these continuous-time systems, quasi-periodicity requires two zero Lyapunov exponents (λ1≈λ2≈0); a positive largest exponent means the window is actually chaotic, while λ1<0 with one zero exponent means periodic. Reporting the spectrum at each θ directly settles whether the claimed quasi-periodic route exists.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The abstract's central route 'chaos → quasi-periodicity → synchronized bursting' depends on a quasi-periodic window that is never directly established. In Section 5, the gap-junction model (3.1) at θ=-1 is labeled quasi-periodic (Fig. 10c) with K=0.3195 and H=0.1827, and the Josephson-junction model (3.5) at θ=-0.5 and -0.1 is labeled quasi-periodic (Fig. 16b,c) with K=0.2207 and K=0.1368. These K values are intermediate: not close to 0 as for the clearly periodic bursting at θ>0 (K≈0.16), nor close to 1 as for the chaotic states. The 0-1 test is designed to separate chaotic (K≈1) from regular (K≈0) dynamics; it does not certify quasi-periodicity, and moderate K values are consistent with weak chaos, intermittency, or a torus. No Lyapunov exponents, Poincaré sections, or frequency spectra are reported for these states. Because the paper's headline finding is a route through quasi-periodicity, this intermediate classification is load-bearing and currently unsupported by any quasi-periodicity-specific diagnostic.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript presents a time-series analysis of small networks of slow-fast denatured Morris–Lecar (dML) neurons under seven coupling strategies (gap junction, thermally sensitive gap junction, chemical, Josephson junction, memristive, higher-order ring-star, and random chemical with autapses). For each model the authors vary the coupling strength (and temperature for the thermally sensitive case) and compute Hurst exponent, sample entropy, the 0–1 test for chaos, Pearson correlation, Kuramoto order parameter, and, for two models, a Granger causality test. The reported phenomenology includes chaos for inhibitory coupling and for elevated temperature, a quasi-periodic regime near zero coupling, synchronized bursting for excitatory coupling, and decay oscillations in some cases. Numerical data and code are made available on GitHub.","tokens_in":40974,"tokens_out":4521,"duration_ms":48354,"significance":"If the reported regime classifications are reliable, the paper provides a useful comparative survey of how popular coupling schemes shape the dynamics of slow-fast bursting neurons, and it doubles as a tutorial for applying several established time-series diagnostics. The strength of the paper is its breadth: seven coupling models, a consistent set of metrics, and openly available code and data, which facilitate reproduction and extension. The central qualitative route 'chaos → quasi-periodicity → synchronized bursting' is, however, supported only by indirect evidence, and the treatment of outlier metric values needs justification; these issues affect the paper's main claims and presently limit its significance.","major_comments":[{"comment":"The quasi-periodic regime, which is load-bearing for the abstract and conclusion, is inferred only from intermediate 0–1 test values and visual inspection. For the gap-junction model at θ = -1 (Fig. 10c) the text states 'Quasiperiodicity is supported by a smaller value of K = 0.3195', and for the Josephson-junction model at θ = -0.5 and -0.1 (Fig. 16b,c) the reported values are K = 0.2207 and K = 0.1368. These values are not close to 0 (regular) or 1 (chaotic), but the 0–1 test only distinguishes chaotic from non-chaotic dynamics; moderate K values are also consistent with weak chaos, intermittency, or a long transient. No Lyapunov exponent, Poincaré section, or frequency spectrum is provided for these states. Please add a quasi-periodicity-specific diagnostic (e.g., Lyapunov spectrum, return map, or spectral peaks) for the states labeled quasi-periodic, or else weaken the claim accordingly.","section":"Section 5 (Eqs. (3.1), (3.5); Figs. 10, 16)"},{"comment":"The post-processing of 'spurious' metric values changes the reported regime boundaries without a validation of the claim that they are artifacts. In Fig. 13, K ≈ -0.48 at (θ, T) ≈ (-2.9, 21.1) is replaced by 0, and several negative H values are described as 'safely ignored'; in Fig. 19, five or six negative K values are replaced by 0. Negative values are not expected for the correlation-based 0–1 test or for the Hurst exponent as defined in Section 4, but discarding them is only legitimate if they can be shown to be algorithmic failures (e.g., by testing on known periodic and chaotic systems with the same time-series length). In addition, all sweeps use a single trajectory per parameter value with randomized initial conditions and no error bars, so sampling variability is unquantified. Please provide robustness checks (multiple initial conditions or noise realizations) and justify the outlier handling, or report the raw values separately.","section":"Section 5 (Figs. 13, 19; text on H values)"},{"comment":"The Granger causality analysis as presented is not a meaningful directional test for the chosen models. For the Josephson junction coupling (3.5) and the memristive coupling (3.6), the couplings are bidirectional and symmetric (each node receives an equal and opposite coupling term), so rejecting the null 'x1 does not Granger-cause x2' is essentially a confirmation that the two time series are coupled, not evidence of a specific direction. Additionally, the text states that for (3.6) the applied strength is θ = 1, but the stated bifurcation range for this model in Table 1 and Section 5 is [-0.02, 0.01]; this appears to be a typo that should be corrected (likely θ = 0.01). Either test a genuinely unidirectional coupling (e.g., chemical coupling (3.4)) or clarify that the test is only detecting coupling, not causal direction.","section":"Section 6 (Granger causality)"}],"minor_comments":[{"comment":"In the definition after \\(\\tilde m(t;e)\\), the growth rate is written as \\(K(e) = \\lim_{t\\to\\infty} \\log m(t;e)/\\log t\\), but the corrected mean-square displacement \\(\\tilde m(t;e)\\) is the quantity used in the regression approach; the formula should refer to \\(\\tilde m(t;e)\\).","section":"Section 4.3 (0–1 test equations)"},{"comment":"The text says 'Here e ∈ (0, 2π) is a small number', but the standard 0–1 test uses a randomly chosen frequency in (0, π) (or (0, 2π)) and does not require e to be small; the wording is misleading.","section":"Section 4.3 (0–1 test parameters)"},{"comment":"The manuscript states that 'the value of H lies in the range [0, 1]', but later reports and discusses negative H values (e.g., §5, Fig. 13 discussion). Please either restrict H to [0, 1] by definition or explicitly state that the R/S estimator can return values outside this range and how such values should be interpreted.","section":"Section 4.1 (Hurst exponent range)"},{"comment":"The caption of Fig. 13 refers to 'model (3.1)', but the figure is for the thermally sensitive gap junction model (3.2); similarly, the caption of Fig. 23 refers to 'model (3.7)', but the figure concerns the random chemical network (3.8). These cross-reference errors should be corrected.","section":"Figure captions (Figs. 13 and 23)"},{"comment":"In the text describing Fig. 11, 'B decreasing monotonically from slightly above 0.9 to beloved 0.8' appears to be a typo for 'below 0.8'.","section":"Section 5 (gap junction sweep)"},{"comment":"In the paragraph after Fig. 16, 'random browninan motion' should be 'random Brownian motion'.","section":"Section 5 (Josephson junction description)"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is a broad phenomenological survey with a tutorial flavor. The main technical gap—the lack of a quasi-periodicity-specific diagnostic—is fixable, and the data/code availability is a strength. I would encourage the editor to request a revision rather than send the paper back for full re-review, provided the authors can either supply the missing diagnostics or appropriately soften the quasi-periodicity claim. The Granger causality section, as written, is more of an illustration than a substantive result, and the reported θ = 1 for model (3.6) should be checked."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a solid, honest numerical atlas for the denatured Morris–Lecar (dML) model under seven coupling strategies, with code on GitHub. It sets out to catalog what standard time-series diagnostics see in small coupled slow–fast networks, and it does that. It is not a methods paper and doesn't resolve a big question, but the coupled dML results for thermal, chemical, Josephson, memristive, higher-order ring-star, and random-with-autapse couplings are new as far as I know. What it does well: clear exposition, tabulated parameters, reproducible scripts, and a consistent use of multiple diagnostics (Hurst, sample entropy, 0–1 test, correlation, Kuramoto factor) across every model. The qualitative picture—chaos for inhibitory coupling, synchronized bursting for excitatory coupling—is consistent and likely correct. The soft spot is the quasi-periodic window. It is central to the abstract's route, but it rests on intermediate 0–1 test values (K ≈ 0.14–0.32) and visual inspection of phase portraits and p–q plots. The 0–1 test separates chaos from non-chaos; it does not certify quasi-periodicity. No Lyapunov exponents, Poincaré sections, or frequency ratios are reported for those states. This is a real gap. Fix: run direct quasi-periodicity diagnostics at the θ ≈ 0 cases, or recast the claim as 'regular but non-periodic' until then. Two smaller issues. First, each parameter point is one trajectory; with random initial conditions, a few seeds per point would make the regime boundaries credible. Second, spurious K and H values are silently trimmed or ignored (K set to 0 or 1, negative H disregarded). That is stated, but not justified. If the algorithm gives impossible values, say why the trimmed value is the right one. Minor, since boundaries probably shift only a little. The Granger causality section is the weakest. Josephson and memristive couplings are symmetric and bidirectional, so a significant Granger test only confirms coupling; it does not show x1 drives x2. The authors essentially say that. Reframe it as a coupling-detection check, or test an asymmetric setup where direction is meaningful. Verdict: send it to external reviewers. Revision should add quasi-periodicity diagnostics and fix the causality interpretation; the rest works as a tutorial reference.","headline":"A reproducible numerical atlas for coupled dML neurons; the quasi-periodicity claim needs direct support and the Granger section is overinterpreted.","tokens_in":41461,"tokens_out":3399,"would_cite":true,"duration_ms":39382,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["37M10","37D45","37N25","92C20"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that sweeping coupling strength in small networks of slow-fast denatured Morris-Lecar neurons moves the dynamics from chaos to quasi-periodicity to synchronized bursting.","keywords":["slow-fast neuron","denatured Morris-Lecar model","bursting dynamics","Hurst exponent","sample entropy","0-1 test","Granger causality","neuronal synchronization"],"falsifier":"Re-run the parameter sweeps with a direct invariant measure, such as the maximal Lyapunov exponent of the full slow-fast system, and with much longer transients; if the regimes with K near 1 and H below 0.5 show no positive Lyapunov exponent, or if the discarded negative K and H values appear at parameter values classed as regular, then the claimed chaos boundaries are artifacts of the 0-1 and Hurst implementations rather than properties of the coupled neuron dynamics.","tokens_in":40500,"feed_emoji":"🧠","tokens_out":6410,"duration_ms":69819,"temperature":0.7,"pith_summary":"The paper asks whether standard time-series diagnostics can tell a coherent story about how coupling strength reshapes the dynamics of small networks of slow-fast bursting neurons. Using the denatured Morris-Lecar model as the node, it simulates seven coupling architectures and sweeps the coupling strength. It reports a repeatable sequence: strong inhibitory coupling yields chaos, near-zero coupling yields quasi-periodicity, and sufficiently excitatory coupling yields phase-synchronized fold/homoclinic bursting. It also finds that thermally sensitive coupling produces chaos only above 20 degrees Celsius, while memristive coupling produces anti-phase decay to equilibrium instead of bursting. The value would be a practical package for classifying real neuron time series.","feed_headline":"Coupling strength steers neurons from chaos to synchrony","feed_subtitle":"Inhibitory links make slow-fast neurons chaotic; excitatory links lock them into phase-synchronized bursts.","key_machinery":"The machinery is a parameter sweep of the coupling strength, plus temperature for the thermal case, across coupled copies of the three-variable slow-fast denatured Morris-Lecar neuron, read through five diagnostics: the Hurst exponent for persistence, sample entropy for irregularity, the 0-1 test for chaos, the Pearson correlation coefficient for in-phase versus anti-phase synchronization, and the Kuramoto order parameter for phase coherence. The 0-1 test is the workhorse that separates chaotic from regular regimes, H and SE corroborate the classification, and Gamma and B identify synchronized bursting.","core_discovery":"The central claim is that the same coupling-strength route appears across most tested architectures: for inhibitory coupling, the 0-1 test returns K near 1, the Hurst exponent drops well below 0.5, and sample entropy rises, identifying chaotic anti-correlated bursting; as the coupling weakens the dynamics become quasi-periodic; for excitatory coupling the nodes lock into fold/homoclinic bursting with H near 0.88, K near 0.16, correlation Gamma equal to 1, and Kuramoto order parameter near 1. The thermally sensitive variant shows chaos only for inhibitory coupling above the reference temperature, while below it the nodes burst anti-phase for inhibition and in-phase for excitation. Chemical coupling stays non-chaotic and asynchronous in the tested range, and memristive coupling gives anti-phase decay to a symmetric equilibrium for excitatory coupling. The paper also claims that Granger causality tests on the Josephson-junction and memristive systems reject non-causality, supporting the interpretation that node 1 drives node 2.","pith_inferences":["The recurrence of the same route across several distinct coupling mechanisms suggests the chaos-to-bursting transition may be a generic feature of slow-fast bursters, which could be tested on FitzHugh-Nagumo or Morris-Lecar neurons with the same diagnostics.","If the diagnostics are trusted, the combination of H and K could classify the sign and strength of synaptic coupling from a single voltage trace: low H with K near 1 points to strong inhibition, while H near 0.88 with K near 0.16 points to excitatory synchronized bursting.","The discarded negative K and negative H values are a testable weak point; recomputing those sweeps with longer runs and a direct Lyapunov exponent would show whether the reported boundaries are real or algorithmic artifacts.","Applying the same package to EEG data, as the authors suggest, assumes the dML time-series statistics transfer to real recordings, and that assumption can be checked by comparing H, SE, and K distributions between model output and recorded signals."],"forward_implications":["With gap-junction coupling, strong inhibition is chaotic and anti-synchronous, while positive coupling gives complete in-phase synchronized bursting.","Thermally sensitive gap junctions switch on chaos only for inhibitory coupling above the reference temperature, so temperature acts as a second bifurcation parameter.","The Josephson-junction strategy reproduces the same chaos-to-quasiperiodicity-to-bursting route, indicating the route is not specific to simple diffusive electrical coupling.","Memristive coupling never becomes chaotic in the tested range and instead ends in anti-phase decay to a symmetric equilibrium point.","The Granger test with lags of at least two rejects non-causality from node 1 to node 2 in the Josephson and memristive two-neuron systems."],"supporting_citations":[{"why":"Supplies the denatured Morris-Lecar model and the slow-fast construction used for every node.","marker":"[Schaeffer and Cain, 2018]"},{"why":"Previous bifurcation analysis of the dML model that guides the parameter values used throughout.","marker":"[Fatoyinbo et al., 2022]"},{"why":"Establishes the single-neuron dML dynamics that serve as the baseline for the coupled systems.","marker":"[Ghosh et al., 2025]"},{"why":"Provides the 0-1 test algorithm and its validity conditions, used to classify chaotic versus regular regimes.","marker":"[Gottwald and Melbourne, 2009a,b, 2016]"},{"why":"Supplies the rescaled-range implementation of the Hurst exponent used for persistence estimates.","marker":"[Qian and Rasheed, 2004]"},{"why":"Provides the sample entropy algorithm used to measure time-series irregularity.","marker":"[Richman and Moorman, 2000]"},{"why":"Defines the chemical coupling scheme with fast threshold modulation used for the non-local synapse model.","marker":"[Belykh et al., 2005]"},{"why":"Introduces the Josephson-junction hybrid synapse coupling adopted for the two-neuron model.","marker":"[Njitacke et al., 2022]"},{"why":"Describes the memristive coupling strategy used for the memristive two-neuron model.","marker":"[Xu et al., 2017]"},{"why":"Models the smallest ring-star network with higher-order interactions, adapted here for dML neurons.","marker":"[Nair et al., 2024]"}],"fun_headline_variants":["Inhibitory chaos, excitatory sync: coupling sign decides","Coupling strength flips neurons between chaos and synchrony","Weak coupling quasi-periodic, strong excitatory synced bursts","Neurons: inhibitory chaos, excitatory sync, weak quasi-periodic","Coupling sign determines neuron dynamics: chaos vs sync"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The 0-1 test and Hurst rescaled-range estimates are assumed reliable on these slow-fast bursting time series, so negative K and negative H values can be dismissed as algorithmic glitches; if those values instead reflect real dynamics, the reported chaos and no-chaos boundaries move.","fun_headline_variants_meta":{"raw":{"variants":["Inhibitory chaos, excitatory sync: coupling sign decides","Coupling strength flips neurons between chaos and synchrony","Weak coupling quasi-periodic, strong excitatory synced bursts","Neurons: inhibitory chaos, excitatory sync, weak quasi-periodic","Coupling sign determines neuron dynamics: chaos vs sync"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001272,"raw_usage":{"total_tokens":5186,"prompt_tokens":908,"completion_tokens":4278,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":524,"completion_tokens_details":{"reasoning_tokens":4192}},"tokens_in":524,"tokens_out":4278,"duration_ms":37537,"temperature":1.0,"reasoning_tokens":4192,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T16:21:22.575640+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-run the parameter sweeps with a direct invariant measure, such as the maximal Lyapunov exponent of the full slow-fast system, and with much longer transients; if the regimes with K near 1 and H below 0.5 show no positive Lyapunov exponent, or if the discarded negative K and H values appear at parameter values classed as regular, then the claimed chaos boundaries are artifacts of the 0-1 and Hurst implementations rather than properties of the coupled neuron dynamics.","supporting_citations":[{"cited_title":"Fatoyinbo, and S.S","cited_arxiv_id":null,"evidence_quote":"Establishes the single-neuron dML dynamics that serve as the baseline for the coupled systems."}],"review_version":1}