{"id":"10a1bf96-0481-413d-9560-a4b4422ec0f3","arxiv_id":"1908.01548","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"In simulated class I and class II Morris-Lecar neurons, ordinal patterns of spike intervals show that signal encoding depends on signal frequency, and electrical coupling transmits the signal more faithfully than chemical coupling.","lead":"This paper uses mathematical symbols for the timing between neuron spikes to see how two Morris-Lecar model neurons, class I and class II, encode and pass along a weak periodic signal. It finds the best encoding and transmission depends on neuron type and whether the connection is electrical or chemical.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Transmission conclusions assume equal ordinal-pattern statistics between sender and receiver imply recoverable signal information; this proxy is not validated, and some regimes infer transmission from permutation entropy alone.","rationale":"The reader's weakest assumption is exactly the load-bearing point: symbol-statistics similarity is used as a proxy for information transmission. In good faith, the paper does define what it means by transmission in Section IV C, and for electrical coupling the full OP distributions are claimed to match, which would imply information transfer if the matching held across all signal parameters. But the evidence presented is qualitative point estimates without error bars, and in the chemical cases the inference sometimes relies on permutation entropy, a scalar that does not determine the distribution. The proposed test directly settles whether the receiver's spike statistics carry signal information, without requiring a specific theoretical commitment. Because this is an addressable evidentiary gap rather than an internal contradiction, the appropriate outcome is the same conditional verdict the reader reached: the central claim should be accepted only after the decoding/mutual-information check is performed. No code, data, or confidence intervals are provided, which reinforces the need for this condition but does not change the verdict category.","tokens_in":10808,"tokens_out":7354,"duration_ms":80820,"concrete_test":"Re-run the unidirectional electrical and class II chemical simulations for A0=250 mV cm2/mS at f=5, 10, 15, 20, 25 Hz with at least 50 independent noise realizations per condition. For each realization, compute the receiver's full 6-bin ordinal-pattern distribution and (i) test whether the distributions differ significantly across f using a permutation test, and (ii) estimate the plug-in mutual information I(f; pattern_receiver) with bias correction. If the receiver's distributions are not significantly different across f, or if I(f; pattern_receiver) is not significantly positive in the regimes where transmission is claimed in Sections IV C and V, then the transmission conclusions fail. Applying the same test to the sender quantifies how much of the sender's encoded information actually reaches the receiver.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central transmission claim depends on the operational definition in Section IV C: a receiver is said to transmit the signal when its ordinal-pattern probabilities and permutation entropy match the sender's. This equates statistical similarity of ISI symbol distributions with information transmission about the signal's amplitude and frequency. A receiver can share the sender's symbol margins, or its entropy, without the signal being decodable from its spike train; this is especially plausible when electrical coupling synchronizes both neurons or when only the scalar permutation entropy is compared (as in the class II chemical unidirectional case discussed in Section V). The paper's own phrase 'in terms of the information defined in this study' (Section IV C) flags the gap. If equality of full receiver and sender OP distributions were established across all f and A0, transmission would follow by a data-processing argument; however, the paper provides point estimates of similarity, no statistical test of equality, and no decoding or mutual-information check. Therefore the abstract's and conclusions' claim that specific class/coupling combinations allow 'more effective transmission of the signal' is not yet supported.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper applies ordinal time-series analysis (ordinal pattern probabilities and permutation entropy) to inter-spike-interval sequences from Morris-Lecar neurons of class I and class II excitability, coupled electrically or chemically, in unidirectional and bidirectional configurations. A weak sinusoidal modulation is applied to one neuron, and the authors study how the signal's amplitude and frequency are encoded in the driven neuron's spike statistics and how these statistics are reflected in the un-driven neuron. The paper reports that the encoding is robust across model variants, reproducing a resonance when the firing rate is twice the signal frequency, and it claims that the best 'transmission' of the signal, defined as similarity of ordinal-pattern statistics between the two neurons, is achieved with electrical coupling, with a class-specific performance for chemical coupling. The conclusions are framed in terms of specific class/coupling combinations allowing 'more effective encoding, or more effective transmission' depending on signal frequency.","tokens_in":11028,"tokens_out":3648,"duration_ms":37412,"significance":"If the operational definition of transmission is accepted, the paper offers a systematic comparison of two excitability classes and two coupling types using a simple symbolic measure, with parameter tables in Appendix A that are sufficiently detailed for re-implementation. The reproduction of the previously reported resonance effect in a different neuron model is a genuine strength, as is the internal consistency of the simulations. However, the central claim about 'more effective transmission' rests entirely on an unvalidated proxy that equates statistical similarity of ordinal patterns with recoverable signal information; this proxy needs to be justified or the claims need to be substantially qualified before the results can be taken as supporting the abstract's conclusions.","major_comments":[{"comment":"The paper defines information transmission as similarity between the sender's and receiver's ordinal-pattern probabilities and permutation entropy, but no decoding or mutual-information measure is used. A receiver can share the sender's symbol margins or entropy without the signal being decodable from its spike train, so the conclusion that specific class/coupling combinations allow 'more effective transmission of the signal' is not supported by the evidence as presented. The phrase 'in terms of the information defined in this study' in Section IV C flags this gap, but the abstract and conclusions do not carry the same caveat. Please either validate the proxy (for example, by showing that OP similarity tracks mutual information between the stimulus and the receiver's ISI sequence, or by demonstrating that the signal can be decoded from the receiver's spikes) or explicitly restrict the claims to 'similarity of ordinal statistics' and revise the title, abstract, and conclusions accordingly.","section":"Section IV C and Section V"},{"comment":"The comparisons of ordinal-pattern probabilities and permutation entropy between sender and receiver are based on point estimates with no error bars, confidence intervals, or statistical tests of equality. The binomial test described in Section III tests whether the probabilities deviate from a uniform distribution; it is not a test of equality between two distributions. Consequently, statements such as 'nearly the same probabilities' (Section V) and the class-specific transmission claims for chemical coupling are not quantitatively supported. Please provide uncertainty quantification or a suitable statistical test for the similarity comparisons.","section":"Section IV C, Figs. 4-5, and Section V"},{"comment":"The conclusion that information is transmitted 'only if the neurons are of class II' for chemical unidirectional coupling is based on a qualitative comparison of entropy maps (Figs. 5(b) and 5(d)), yet Section IV C itself states that the receiver's entropy is higher than the sender's for both classes. Since permutation entropy alone measures orderliness, not transmitted signal information, this specific claim is not established by the presented evidence. Either provide a more direct measure of transmitted information for this case or rephrase the conclusion to describe the observed entropy similarity rather than information transmission.","section":"Section V, class II chemical unidirectional case"}],"minor_comments":[{"comment":"The word 'Plato' should be 'plateau' in the sentence describing the receiver neuron's entropy behavior.","section":"Section IV B"},{"comment":"The verb 'codify' should be 'encode' in the sentence about the two coupled neurons as a single unit.","section":"Section V"},{"comment":"The text uses 'Class 1' and 'Class 2' in one sentence while the rest of the paper uses 'class I' and 'class II'; please standardize the notation.","section":"Appendix A"},{"comment":"Reference 23 lists the year as 2016 for a paper in Phys. Rev. E 79, but that volume corresponds to 2009; please correct the date. Also, Reference 24 abbreviates 'Comput Biol Med' as 'Compt Biol Med'; please fix the journal title.","section":"References"},{"comment":"The subscript for the fast conductance appears as 'g f ast' with a space; please correct the typo to 'g_fast'.","section":"Equation (1)"}],"recommendation":"major_revision","confidential_remarks":"This is a competent computational study that extends the authors' prior work on ordinal analysis of neuron models. The main technical concern is the unvalidated definition of 'information transmission' as statistical similarity of ordinal patterns and entropy; this is a load-bearing issue for the central claim. The paper would be acceptable if the authors either validate the proxy with a decoding-based measure or significantly soften the claims and title. The simulations and parameter documentation are sufficient for reproducibility, which counts in the paper's favor."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Candid take: this is a competent, clearly written simulation study that extends the group's ordinal-analysis work from FitzHugh-Nagumo to Morris-Lecar class I and class II neurons, adding a systematic comparison of chemical vs electrical coupling in unidirectional and bidirectional configurations. The parameter tables in Appendix A are complete enough to re-implement the model, and the simulations look internally consistent. The resonance at twice the firing rate is reproduced from prior work, not assumed, so the core computations stand on their own.\n\nWhat is genuinely new is the comparative map: how ordinal-pattern statistics of ISIs depend on excitability class and coupling type. That is useful for anyone choosing between analysis methods for spike trains, and it goes beyond an off-the-shelf application. The paper shows which regimes produce distinct pattern statistics and which couplings preserve them.\n\nThe soft spot is the transmission claim. The paper defines transmission as similarity of ordinal-pattern probabilities and permutation entropy between sender and receiver (Section IV C), then says electrical coupling transmits better because the receiver's statistics look like the sender's. That is a proxy, not a measure of recoverable information. The authors do flag it—\"in terms of the information defined in this study\"—but the abstract and conclusions drop the qualification, so the headline finding overstates. If the full symbol distributions were equal, a data-processing argument would give genuine preservation of signal information; but the paper shows point estimates only, with no statistical test of equality, and some regimes compare just the scalar entropy. That is fixable in revision: report distributional distances with confidence intervals, or use a decoding/transfer-entropy measure.\n\nMinor concerns: coupling conductances are hand-fitted to reach 1:1 locking, noise intensities are chosen to match firing rates, and there are no error bars or shared code/data. These are not fatal for a modeling paper, but they limit how strongly the comparative rankings can be read. The citation pattern is fine—the same group wrote the earlier FHN papers, but these Morris-Lecar simulations are independent, so this is not a self-citation problem.\n\nBottom line: the paper deserves a serious referee. It is a solid, re-implementable simulation study with a clear scope, and the transmission proxy is addressable. I'd want either a stronger transmission analysis or softened wording. For a reading group, it is a good example of ordinal-pattern methods on coupled conductance-based neurons; I would cite it if I were working on ordinal methods for spike trains.","headline":"A solid, re-implementable simulation study whose comparative encoding results hold up, but whose transmission conclusions overstate a similarity proxy.","tokens_in":11548,"tokens_out":3883,"would_cite":true,"duration_ms":36778,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Weak signals leave class- and coupling-dependent fingerprints in spike-interval patterns.","keywords":["ordinal analysis","permutation entropy","Morris-Lecar model","class I excitability","class II excitability","signal encoding","information transmission","electrical synapse"],"falsifier":"Run the same coupled Morris-Lecar simulations and compute the mutual information between the sinusoidal input and the follower neuron's spike train (or decode the signal from the follower's ISIs) at the parameter combinations flagged as best transmission, such as electrical coupling at $f = 20$ Hz for class II. If the follower's ordinal probabilities match the sender's but mutual information is at chance level, the paper's transmission measure would be shown to be a synchrony artifact rather than information transfer.","tokens_in":10585,"feed_emoji":"🧠","tokens_out":5296,"duration_ms":54063,"temperature":0.7,"pith_summary":"This paper asks whether the statistics of spike-timing patterns can reveal how a weak periodic signal is encoded and transmitted between two coupled neurons. Using ordinal analysis, which turns sequences of inter-spike intervals into a small set of symbols and counts how often each symbol appears, the authors compare a neuron that receives a weak periodic input with a follower neuron that does not. They study two excitability classes of the Morris-Lecar model, class I and class II, under electrical and excitatory chemical couplings, in both one-way and two-way arrangements. The central claim is that the choice of neuron class and coupling type matters in a frequency-dependent way: some combinations encode the signal better, while others transmit it more faithfully. If correct, ordinal-pattern statistics provide a sensitive symbolic readout of signal features that conventional linear measures miss.","feed_headline":"Electrical coupling transmits weak signals best in both neuron classes","feed_subtitle":"Spike-interval patterns carry amplitude and frequency information; class I and class II split slow and fast encoding.","key_machinery":"The operative tool is ordinal time-series analysis applied to inter-spike intervals. Each ISI is compared with the two that follow it (embedding dimension $L=3$), producing one of $3! = 6$ ordinal patterns; the paper tracks pattern probabilities and the permutation entropy $H$, normalized so $H=1$ means all patterns are equally likely and smaller $H$ means that preferred spike-timing orders stand out. This symbolic reduction is what lets the authors quantify encoding (the spread and stability of pattern probabilities as amplitude and frequency vary) and transmission (how closely the follower neuron's pattern probabilities and entropy match the sender's). A key supporting observation is a resonance: the probability of pattern 210 dips when the firing rate is near twice the modulation frequency, for both neuron classes and both coupling types.","core_discovery":"The paper's central discovery is that ordinal-pattern statistics of inter-spike-interval sequences carry frequency- and amplitude-specific information about a weak periodic input, and that encoding and transmission do not go together. In unidirectionally coupled neurons, class I neurons resolve low signal frequencies better and class II neurons resolve high frequencies better. Electrical coupling, by synchronizing the two neurons, transmits the symbolic pattern statistics almost unchanged to the follower neuron for both classes, whereas unidirectional chemical coupling transmits only in the class II case. With bidirectional chemical coupling, the pair behaves as a single processing unit: class I neurons can encode a wider range of frequencies than either neuron alone, but the follower no longer reproduces the sender's patterns, so transmission in the authors' sense fails. The same signal features are therefore not simultaneously optimized; the better transmission route (electrical) and the better encoding route (bidirectional chemical, class I) are different.","pith_inferences":["Beyond the paper, the transmission ranking could be re-tested with mutual information between the stimulus and the follower's spike train, since similar ordinal statistics do not by themselves guarantee that the signal is recoverable.","Beyond the paper, the predicted frequency split (class I better at slow, class II at fast) could be tested by driving regular-spiking and fast-spiking neurons in vitro with sinusoidal currents and applying ordinal analysis to the recorded spike trains.","Beyond the paper, because ordinal patterns of length three ignore interval durations, combining pattern frequencies with ISI averages would likely separate amplitude encoding from frequency encoding more sharply."],"forward_implications":["Class I and class II neurons are complementary rather than redundant encoders: slow signals are best read from class I spike trains, fast signals from class II, so a population containing both classes could cover a wider frequency band.","Electrical (gap-junction) coupling is the reliable transmission channel in these simulations because it forces the follower neuron's ordinal statistics to match the sender's, making diffusive coupling a candidate substrate for carrying periodic stimulus features across a network.","Bidirectional chemical coupling does not transmit the signal in the symbolic sense, but it expands the encoding range of class I neurons; mutually coupled pairs may act as a single encoding unit rather than a sender-receiver link.","The encoding resonance at firing rate equal to twice the signal frequency gives a concrete prediction: a neuron's preferred symbolic pattern should respond non-monotonically to changes in baseline firing rate, with strongest ordering near the resonance."],"supporting_citations":[{"why":"Introduced ordinal-pattern probabilities as a way to read the period and amplitude of a weak signal from inter-spike-interval sequences in a noisy neuron.","marker":"[8]"},{"why":"Showed that this ordinal encoding survives when the modulated neuron is coupled to a second neuron, the setup this paper extends to transmission.","marker":"[15]"},{"why":"Supplies the Morris-Lecar model whose two parameter regimes produce the class I and class II excitability compared here.","marker":"[16]"},{"why":"Provides the specific two-dimensional Morris-Lecar formulation and parameter set the simulations use.","marker":"[18]"},{"why":"Defines class I and class II excitability through the continuity of the frequency-current curve, the classification that organizes the results.","marker":"[17]"},{"why":"Defines permutation entropy, the statistic used to quantify how ordered (non-uniform) the ordinal pattern distribution is.","marker":"[21]"}],"fun_headline_variants":["Encoding and transmission split: electrical transmits, chemical encodes","Weak signal handling differs by neuron class and coupling type","Spike patterns reveal trade-off between encoding and transmission","Class I for slow, class II for fast — coupling affects transmission"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The paper treats a follower neuron as transmitting the signal when its ordinal-pattern probabilities and permutation entropy match the sender's, without any decoding step or mutual-information measure that would show the signal can be recovered from the follower's spike train.","fun_headline_variants_meta":{"raw":{"variants":["Encoding and transmission split: electrical transmits, chemical encodes","Weak signal handling differs by neuron class and coupling type","Spike patterns reveal trade-off between encoding and transmission","Class I for slow, class II for fast — coupling affects transmission"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000389,"raw_usage":{"total_tokens":2066,"prompt_tokens":974,"completion_tokens":1092,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":590,"completion_tokens_details":{"reasoning_tokens":1026}},"tokens_in":590,"tokens_out":1092,"duration_ms":11831,"temperature":1.0,"reasoning_tokens":1026,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:09:31.046295+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same coupled Morris-Lecar simulations and compute the mutual information between the sinusoidal input and the follower neuron's spike train (or decode the signal from the follower's ISIs) at the parameter combinations flagged as best transmission, such as electrical coupling at $f = 20$ Hz for class II. If the follower's ordinal probabilities match the sender's but mutual information is at chance level, the paper's transmission measure would be shown to be a synchrony artifact rather than information transfer.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduced ordinal-pattern probabilities as a way to read the period and amplitude of a weak signal from inter-spike-interval sequences in a noisy neuron."},{"cited_title":"& author Masoller, C","cited_arxiv_id":null,"evidence_quote":"Showed that this ordinal encoding survives when the modulated neuron is coupled to a second neuron, the setup this paper extends to transmission."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Morris-Lecar model whose two parameter regimes produce the class I and class II excitability compared here."},{"cited_title":"A., De Koninck, Y., Sejnowski, T","cited_arxiv_id":null,"evidence_quote":"Provides the specific two-dimensional Morris-Lecar formulation and parameter set the simulations use."},{"cited_title":"M., Dynamical Systems in Neuroscience: The Geometry of Excitability and Bursting The MIT Press, Cambridge, Massachusetts, London, England, 2010","cited_arxiv_id":null,"evidence_quote":"Defines class I and class II excitability through the continuity of the frequency-current curve, the classification that organizes the results."},{"cited_title":"& author Pompe, B","cited_arxiv_id":null,"evidence_quote":"Defines permutation entropy, the statistic used to quantify how ordered (non-uniform) the ordinal pattern distribution is."}],"review_version":1}