{"id":"a22e021e-fd81-4f76-804b-29172b27060f","arxiv_id":"2608.12208","paper_version":1,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Multi-state switching environments broaden the population size distribution and change the slow strain's fixation probability relative to binary switching, with the direction depending on switching rate, bias, and the median carrying capacity.","lead":"Microbial strains competing in environments that switch gradually through several resource levels are studied with new stochastic models. The work shows that adding intermediate environmental states can either help or hurt the slower strain's fixation odds, depending on the resource levels and switching speed.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Binary comparison depends on matching only mean transition times; alternative calibration by autocorrelation time could shift the quantitative boundaries of φ_{2n+1} vs φ_2, though the core multi-state results are unaffected.","rationale":"The paper's central contribution is a careful characterization of multi-state switching dynamics: the PSD, the fixation probability, and the uMFT, together with an N-PDMP approximation validated by extensive Gillespie simulations. The strongest claim, as formulated by the reader, involves a comparison to a 'matched binary environment', and the matching convention in Eq. (8) is the least secure part of that claim. The multi-state transition-time distribution is Erlang-like, not exponential, so matching only the first moment is an arbitrary calibration. A different convention, such as matching autocorrelation time or carrying-capacity variance, will change the effective binary switching rates and hence the quantitative boundaries between regions where φ_{2n+1} > φ_2 and φ_{2n+1} < φ_2. This is a genuine limitation, but it is explicitly stated in the paper and is a modeling choice rather than an internal inconsistency. The qualitative picture—that low K_0 makes multi-state environments harsher and enhances slow-strain fixation relative to binary, while high K_0 suppresses it—is robust because it follows from the stationary distribution of K and the monotonicity of the Moran fixation probability. The N-PDMP approximation (Eq. 19) is independently supported by the strong agreement with Gillespie simulations in Figs. 5 and 6, and the derivations of the quenched/annealed limits (Eqs. 16–17) are standard. I therefore agree with the reader's identification of the weakest assumption and do not see a reason to change the ACCEPT verdict; a robustness check with an alternative calibration would strengthen the comparison but is not essential for the multi-state results.","tokens_in":36437,"tokens_out":24324,"duration_ms":196881,"concrete_test":"Recompute the effective binary fixation probability φ_2 for the parameter sets of Figs. 6(a)-(f) using eν_alt = -λ1/2, where λ1 is the largest non-zero eigenvalue of the (2n+1)-state generator Q, instead of Eq. (8); record whether the sign of φ_5 - φ_2 changes at any ν_5 in those panels and whether the crossover values of K_0 quoted in Sec. III B 4 shift by more than 20%.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Equation (8) calibrates the effective binary model to the multi-state model by matching the mean transition times of K−→K+ and K+→K−. This is the only matching constraint; it does not match the variance or the autocorrelation time of the environmental noise. In the multi-state model, the dwell time from K− to K+ is the sum of 2n exponential waiting times (Erlang), whereas the binary model has exponential dwell times. For example, for n=2, ϵ=0, δ=0, the 5-state process has autocorrelation time t_c ≈ 2.618/ν (largest non-zero generator eigenvalue −0.382ν), while the binary rate from Eq. (8), eν = ν/4, gives t_c = 1/(2eν) = 2/ν. Matching t_c instead would require eν ≈ 0.191ν, a 24% change. Because the binary curves in Figs. 5–6 are plotted against ν_{2n+1} through eν(ν_{2n+1}), such a change shifts the binary fixation probability curves horizontally. The reported thresholds such as 'φ_3^0 > φ_2^0 for K_0 ≲ 141.36' (Sec. III B 3) and the regimes where φ_5 and φ_2 cross (Fig. 6(e)) are therefore specific to the mean-transition-time convention. The qualitative statements that low K_0 enhances and high K_0 suppresses φ_{2n+1} relative to φ_2 are likely robust because they follow from the presence of low/high intermediate carrying capacities, but the quantitative comparison is convention-dependent. This is explicitly acknowledged as a modeling choice, not an error, and it does not affect the multi-state results or the validation of Eq. (19).","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies a well-mixed population of two competing strains, a slow-growing S and a faster F, under a carrying capacity K(t) that switches among 2n+1 environmental states with linearly spaced values K_i. The environment is a Markov chain with nearest-neighbour switching, characterized by rates ν_{2n+1}, bias δ, and asymmetry ε. Using 10^5-realization Gillespie simulations and an N-PDMP approximation that averages the Moran fixation probability over the PDMP stationary density with a rescaled switching rate (Eq. 19), the authors compute the population size distribution, average population size, fixation probability of S, and unconditional mean fixation time. They compare these quantities with an effective binary switching model calibrated by matching mean transition times (Eq. 8). The central findings are that multi-state environments produce a broader PSD, that the slow strain's fixation probability can be higher or lower than in the matched binary environment depending on K_0, δ, n, and ν, and that the N-PDMP approximation accurately captures the fixation probability across slow, intermediate, and fast switching (Figs. 5 and 6).","tokens_in":36798,"tokens_out":4945,"duration_ms":50337,"significance":"If the comparative claims are robust, this is a valuable and experimentally motivated extension of binary switching models to multi-state feast-famine cycles. The paper's strengths are its systematic validation against large-scale Gillespie simulations, the reusable N-PDMP approximation of Eq. (19) for fixation probability and mean fixation time, the analytical slow/fast switching limits in Eqs. (16) and (17), and the open data and code. The main reservation concerns the calibration convention used for the binary comparator, which affects the quantitative comparisons but not the internal validation of the multi-state results.","major_comments":[{"comment":"The effective binary rates in Eq. (8) are defined by matching only the mean transition times between K− and K+, but the multi-state dwell time from famine to feast is a sum of n exponential waiting times (Erlang) whereas the binary model has exponential dwell times. Since the binary curves in Figs. 5 and 6 are plotted against ν_{2n+1} through eν(ν_{2n+1}), an alternative calibration—for example matching the autocorrelation time or the variance of K—shifts the binary curves horizontally. For the symmetric 5-state case the autocorrelation time changes eν by about 24% relative to Eq. (8). This convention-dependence directly affects quantitative statements such as the crossing between φ_5 and φ_2 in Fig. 6(e) and the location of intermediate-regime thresholds. The manuscript states the matching convention but does not test its robustness. I ask the authors to add a short analysis (or an explicit caveat) demonstrating that the qualitative conclusions of Sec. III B are insensitive to physically reasonable alternative matching rules, or to reframe the comparative claims as convention-specific.","section":"II A, Eq. (8); III B 2 and Fig. 6(e)"}],"minor_comments":[{"comment":"The caption states 'Here, n=1' for the five-state model; this should read n=2.","section":"Fig. 8 caption"},{"comment":"The captions refer to the N-PDMP approximation 'see Eq. (19)' for the mean fixation time; the correct reference is Eq. (B2).","section":"Figs. 7 and 8 captions"},{"comment":"The text mentions '2nenvironmental states' as a possible extension; this is presumably a typo for '(2n+1)-state' or '2n+1 environmental states'.","section":"IV, extension paragraph"},{"comment":"There is a duplicated phrase 'the number of the number of intermediate states' in the sentence explaining the slower convergence for n=2.","section":"III A, near Eq. (15)"},{"comment":"The asserted insensitivity of PSD, φ_{2n+1}, and τ_{2n+1} to ε is supported by a small set of parameter combinations and no error analysis; a sentence explaining why the ε-dependence cancels (or a systematic scan) would make the claim more convincing.","section":"Appendix D"}],"recommendation":"major_revision","confidential_remarks":"This is a high-quality technical contribution with strong numerical validation and open code/data. My recommendation is driven by a single robustness issue: the comparison with the binary model depends on the mean-transition-time calibration of Eq. (8), and the manuscript would be strengthened by a robustness check or by explicitly limiting the quantitative claims to that convention. The remaining items are presentation-level."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper does exactly what it claims: generalizes the binary switching carrying-capacity model to (2n+1) ordered states and gives a systematic way to compute population-size statistics and fixation probabilities in that setting. The new pieces are the linearly spaced carrying-capacity construction, the effective binary rate matching (Eq. 8), the N-PDMP approximation (Eq. 19), and the explicit n=1,2 results showing that intermediate states can either enhance or suppress the slow strain's fixation probability relative to a matched binary environment. The analytical approximations are checked against 10^5-realization Gillespie simulations across a wide parameter range, and the agreement in Figs. 5 and 6 is genuinely good. The paper also ships code and data, which is real evidence and should count in its favor.\n\nThe main soft spot, which the reader's stress-test correctly identifies, is that the multi-state vs. binary comparison is calibrated by matching only the mean transition times between extremes. The multi-state dwell time is Erlang, not exponential, so the binary telegraph noise has a different correlation structure. Matching autocorrelation time instead would change the effective binary rate by roughly 24% in the n=2, δ=0 case, and would shift the quantitative thresholds like K0≈141.36 in Sec. III B 3. This does not undermine the multi-state results themselves—the PSD, fixation probability, and N-PDMP validation stand on their own—but the specific claims about where φ_{2n+1} is above or below φ_2 are convention-dependent. The author states the convention clearly, so it is a modeling choice rather than an error, but the reader should not treat the quantitative boundaries as robust predictions about feast-famine cycles in nature.\n\nMinor issues, in proportion: the claimed insensitivity to ε and the inequality K_{2n+1}>K_2 for δ>0 are supported numerically rather than analytically, which is fine but should be flagged as such. There is a typo in Eq. (B2) where the fast-switching line says ν≪s instead of ν≫s; easy to fix. The paper builds heavily on the author's own prior binary results, but the central comparison is against independent simulations, so I do not see circularity as a real problem.\n\nWho is this for? Someone working on eco-evolutionary dynamics in fluctuating environments, especially if they care about coarse-grained feast-famine cycles with intermediate states. It is a solid extension within an established research program, not a revolution. I would be happy to referee it; the analysis is careful, the numerics are extensive, and the few soft spots are explicitly acknowledged or easily patched.","headline":"Solid, careful extension of the binary-switching program to multi-state feast-famine environments; the core multi-state results are well supported by simulations, while the headline comparison to binary environments is tied to one matching convention.","tokens_in":37330,"tokens_out":825,"would_cite":true,"duration_ms":9754,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["92D15","92D25","60J27"],"pacs":[],"model":"deepseek-v4-flash","headline":"In feast-famine cycles with intermediate resource levels, a slow strain's fixation probability can rise above or drop below its value under matched binary switching, and a piecewise-deterministic Markov process approximation predicts the…","keywords":["multi-state switching environments","feast-famine cycles","fixation probability","stochastic population dynamics","carrying capacity fluctuations","piecewise-deterministic Markov process","Moran model","environmental noise"],"falsifier":"Recompute the multi-state and binary fixation probabilities with the binary environment calibrated to match the autocorrelation time or the stationary variance of the carrying capacity instead of the mean transition time of Eq. (8), and check whether the regions where multi-state fixation exceeds binary fixation, such as low median carrying capacity, persist; if they vanish or flip sign, the reported enhancement of slow-strain fixation in gradual environments is an artifact of the mean-time calibration convention.","tokens_in":36196,"feed_emoji":"🦠","tokens_out":9884,"duration_ms":79888,"temperature":0.7,"pith_summary":"The paper asks whether a feast-famine environment with several intermediate resource levels leads to different evolutionary outcomes than the usual two-state (feast/famine) model. It claims that adding intermediate states, modeled as a (2n+1)-state Markov chain of carrying capacities, changes the slow strain's fixation probability in a parameter-dependent way: with a low median carrying capacity K_0, the slow strain fixates more often in the multi-state environment than in a properly matched binary one; with a high K_0, the opposite holds, and the switching bias shifts the crossover. Why it matters: if true, the standard binary coarse-graining of nutrient fluctuations can systematically misestimate fixation and mean fixation time in real feast-famine cycles, and the paper supplies a fast approximation that predicts these quantities without full simulation.","feed_headline":"Intermediate states can raise or lower slow-strain fixation","feed_subtitle":"A piecewise-deterministic approximation predicts exactly when gradual environments favor the slower competitor.","key_machinery":"The central object is the (2n+1)-state switching model, in which the carrying capacity moves up and down a ladder of evenly spaced values between famine and feast, driven by a continuous-time Markov chain. The key approximation is the N-PDMP: a piecewise-deterministic Markov process in which the population size evolves deterministically by a logistic equation between random environmental switches, and whose stationary density is used to average the Moran fixation probability. The load-bearing identity is the mean-transition-time matching condition of Eq. (8), which fixes effective binary switching rates so that the multi-state and binary environments have the same average famine-to-feast and feast-to-famine travel times; this calibration makes the comparison between multi-state and binary fixation probabilities meaningful. The approximation works by replacing the intractable individual-based process with the deterministic-plus-random N-PDMP, then averaging the Moran formula over its stationary density with a rescaled switching rate.","core_discovery":"In multi-state switching environments, the fixation probability of the slower strain is not simply a smoothed version of the binary result. The paper shows that for a (2n+1)-state environment, the slow-strain fixation probability can be larger or smaller than its binary counterpart depending on the median carrying capacity, the environmental bias, and the switching rate, and that the piecewise-deterministic-Markov-process (N-PDMP) approximation quantitatively captures it across slow, intermediate, and fast switching in three- and five-state models. In the slow-switching limit, the fixation probability is a weighted average of Moran fixation probabilities over the 2n+1 carrying capacities; in the fast-switching limit it is governed by an effective carrying capacity that is the harmonic mean of the K_i over the stationary distribution. Intermediate-state effects are strongest when the median carrying capacity is low, because the population then spends enough time at small sizes that demographic noise boosts the slow strain's chances, and they are weaker or reversed when the median carrying capacity is high.","pith_inferences":["If a different 'same environment' calibration is used, such as matching the autocorrelation time or the stationary variance of the carrying capacity instead of the mean famine-to-feast transition time of Eq. (8), the boundaries of the regions where multi-state fixation exceeds binary fixation, including the low-median-capacity enhancement regions, may shift or disappear.","The mechanism behind the low-median-capacity enhancement, namely repeated population bottlenecks at intermediate carrying capacities amplifying demographic noise, predicts that a chemostat with a gradual nutrient ramp should fixate the slower strain more often than an abrupt two-state chemostat with the same extreme concentrations and the same mean cycle period, which is a directly testable experi","Taking the number of intermediate states to infinity with suitably scaled rates would turn the discrete ladder of carrying capacities into a continuum, providing a bridge to established continuous-environmental-noise models and a test of whether the discrete-state effects reported here survive in the continuum limit."],"forward_implications":["If the N-PDMP approximation of Eq. (19) retains its accuracy for more than two intermediate states, fixation probabilities and mean fixation times in environments with many states can be computed without full individual-based simulation, at a fraction of the cost.","The sign of the difference between multi-state and binary fixation probabilities is not fixed: a low median carrying capacity makes the slow strain more likely to fixate in multi-state environments, while a high median carrying capacity makes it less likely, and the environmental bias shifts the crossover point.","Increasing the number of intermediate states generally lowers the slow strain's fixation probability at fixed median carrying capacity, but parameter sets exist where the multi-state environment beats the binary one in some switching regimes and loses in others, so binary-model conclusions can miss regime-dependent reversals.","The unconditional mean fixation time scales with the inverse selection strength, with a prefactor set by the number of states, the environmental bias, and the median carrying capacity; the N-PDMP approximation captures its dependence on the switching rate, giving direct predictions for time-to-fixation in gradual environments."],"supporting_citations":[{"why":"Supplies the Moran-model fixation probability formula used in the slow- and fast-switching averages and in the N-PDMP approximation.","marker":"[8]"},{"why":"Introduces the piecewise-deterministic Markov process approximation for population size under binary environmental switching, which this paper generalizes to multiple states.","marker":"[18]"},{"why":"Provides the detailed binary switching eco-evolutionary model that serves as the baseline to which multi-state fixation probabilities are compared.","marker":"[19]"},{"why":"Establishes the equivalence between random and periodic binary switching and the effective-rate matching convention that underlies Eq. (8).","marker":"[21]"},{"why":"Experimental evidence that feast-famine cycle frequency and amplitude shape microbial population dynamics, motivating the multi-state extension.","marker":"[48]"},{"why":"Shows intermediate nutrient conditions naturally arise in feast-famine dynamics, motivating the inclusion of intermediate carrying-capacity states.","marker":"[49]"},{"why":"Experimental comparison of binary and ternary environments demonstrating differences between abrupt and gradual environmental change, motivating the model.","marker":"[51]"}],"fun_headline_variants":["Multi-state switching can boost or cut slow-strain fixation","Slow-strain fixation depends on median carrying capacity","Fast-switch effective capacity is harmonic mean of states","Slow-strain fixation can exceed or fall below binary case"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The comparison between multi-state and binary environments assumes that matching only the average time for the carrying capacity to move from famine to feast makes the two environments equivalent, even though the multi-state dwell time is a sum of several exponential waiting times and is not exponentially distributed.","fun_headline_variants_meta":{"raw":{"variants":["Multi-state switching can boost or cut slow-strain fixation","Slow-strain fixation depends on median carrying capacity","Fast-switch effective capacity is harmonic mean of states","Slow-strain fixation can exceed or fall below binary case"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000689,"raw_usage":{"total_tokens":3121,"prompt_tokens":945,"completion_tokens":2176,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":561,"completion_tokens_details":{"reasoning_tokens":2112}},"tokens_in":561,"tokens_out":2176,"duration_ms":15073,"temperature":1.0,"reasoning_tokens":2112,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-16T00:12:30.814146+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Recompute the multi-state and binary fixation probabilities with the binary environment calibrated to match the autocorrelation time or the stationary variance of the carrying capacity instead of the mean transition time of Eq. (8), and check whether the regions where multi-state fixation exceeds binary fixation, such as low median carrying capacity, persist; if they vanish or flip sign, the reported enhancement of slow-strain fixation in gradual environments is an artifact of the mean-time calibration convention.","supporting_citations":[{"cited_title":"Abdul-Rahman, D","cited_arxiv_id":null,"evidence_quote":"Introduces the piecewise-deterministic Markov process approximation for population size under binary environmental switching, which this paper generalizes to multiple states."},{"cited_title":"Ewens.Mathematical Population Genetics","cited_arxiv_id":null,"evidence_quote":"Provides the detailed binary switching eco-evolutionary model that serves as the baseline to which multi-state fixation probabilities are compared."},{"cited_title":"Dobramysl, M","cited_arxiv_id":null,"evidence_quote":"Establishes the equivalence between random and periodic binary switching and the effective-rate matching convention that underlies Eq. (8)."},{"cited_title":"Szolnoki and M","cited_arxiv_id":null,"evidence_quote":"Experimental evidence that feast-famine cycle frequency and amplitude shape microbial population dynamics, motivating the multi-state extension."},{"cited_title":"Esmaeili, B","cited_arxiv_id":null,"evidence_quote":"Shows intermediate nutrient conditions naturally arise in feast-famine dynamics, motivating the inclusion of intermediate carrying-capacity states."},{"cited_title":"Kussell, R","cited_arxiv_id":null,"evidence_quote":"Experimental comparison of binary and ternary environments demonstrating differences between abrupt and gradual environmental change, motivating the model."}],"review_version":1}