{"id":"bfdb844c-d910-4023-afa5-14e44af3a312","arxiv_id":"2507.10332","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":9,"one_line_summary":"Under sex-dependent continuation probabilities, the evolutionary stable strategy is a 1:1 sex ratio among last-born children, not necessarily at birth, offering a possible explanation for the male-biased human birth ratio.","lead":"A game-theory model shows that when parents are more likely to have another child after a son than after a daughter, evolution can favor a male-biased sex ratio at birth, but the sex ratio among last-born children stays 1:1. The paper uses this to suggest why humans are born about 105 boys per 100 girls.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"ESS proof depends on the non-standard matched-pair fitness C^B (Eq. 7); the paper's own Fig. 3 shows p*_s is not an ESS under the standard whole-population measure at g=2, so the stability claim is not established under accepted invasion fitness.","rationale":"The reader's verdict identified the same weakest assumption: the ESS proof rests on the new matched-pair fitness C^B. I agree. The paper's most valuable, self-contained result is the deterministic steady-state formula Eq. (2) and the data collapse in Fig. 1(c), which is supported by the simulations and does not depend on the fitness measure. The SRLB=1:1 result is a mathematical consequence of Eq. (4) once the population is at p*_s. The weak link is the claim that p*_s is an ESS: the authors cannot prove this under the standard whole-population gene-frequency measure at the second generation (Fig. 3), and their positive proof uses a non-standard measure whose biological status is not justified. The Mathematica proof in Appendix B is also not independently checkable (no code released). I do not regard this as a fundamental error—the long-run simulations suggest p*_s may well be stable under the standard measure—but the paper overclaims when it says it mathematically proves ESS. A conditional verdict, requiring the authors to connect C^B to invasion fitness or to provide the standard-metric proof, is appropriate. The human application's discrepancy (103 vs 105) is secondary and not the load-bearing issue.","tokens_in":11666,"tokens_out":7331,"duration_ms":88660,"concrete_test":"Perform a rare-mutant invasion analysis under the standard whole-population fitness: linearize the recursion Eq. (5) at the resident equilibrium with all individuals carrying p*_s, and compute the leading eigenvalue (or simulate the frequency of a small mutant cohort over many generations). If the mutant frequency (X^B) declines for every nonzero Delta, then p*_s is a genuine ESS and the C^B-based proof is corroborated. If, as Fig. 3 suggests at g=2, a nonzero Delta initially increases, the ESS claim is an artifact of the C^B definition and the central claim must be revised or the measure justified.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that the 1:1 SRLB strategy p*_s is evolutionarily stable is proven only with respect to the newly introduced fitness measure C^B, the gene frequency in matched pairs (Eq. 7). This measure is not the standard invasion fitness for a rare mutant; the standard measure is the mutant's contribution to the gene pool of future generations, which the authors themselves track via X^B_g (Eq. 6). Fig. 3 shows that at g=2 (the conventional ESS check) X^B_g increases monotonically with Delta for Delta != 0, so p*_s is not an ESS under that standard measure. The authors argue that at much later generations (e.g., g=10^4) X^B_g is maximized at Delta=0, but this is only a numerical observation, and the choice of generation is arbitrary. The formal Mathematica proof in Appendix B only shows that p*_s maximizes C^B; it does not show that maximizing C^B is equivalent to maximizing long-term gene-frequency growth, nor that C^B is the correct fitness currency for sex-ratio evolution. Without such an equivalence, the proof relocates the target: it defines a fitness measure under which the result holds, rather than showing the strategy is uninvadable under the standard population-genetic definition. The human application inherits this fragility: the fitted p*_s=0.512 is claimed stable because of the C^B-based proof, but the year-by-year simulation reaches only 103:100, and the authors speculate about missing mortality details. The simulations do independently support that the deterministic steady state obeys Eq. (3) and that SRLB converges to 1:1; that part is not in question. What is load-bearing is the ESS claim, and that rests entirely on the untested fitness measure.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies a sex-ratio evolution model in which a couple's probability of having another child depends on the sex of the previous child, with continuation probabilities bs and bd. The authors derive a steady-state son probability p*_s = (1-bd)/(2-bs-bd) and show algebraically (Eq. 4) that this condition is equivalent to a 1:1 sex ratio among last-born children (SRLB), even when the sex ratio at birth is biased. They claim that this 1:1 SRLB is an evolutionarily stable strategy under a newly introduced fitness measure, the frequency of a gene in matched reproductive pairs (C^B), and they apply the model to the human male-biased SRB by fitting bs and bd to historical mortality and the observed 105:100 sex ratio. The steady-state formula is supported by a data collapse across 100 parameter combinations (Fig. 1c) and by simulations showing convergence of the last-born sex ratio to 0.5 (Fig. 2). The ESS proof, however, relies on a nonstandard fitness measure and a Mathematica computation in Appendix B that is not shown.","tokens_in":11985,"tokens_out":12356,"duration_ms":140022,"significance":"If confirmed, the paper would offer a novel, falsifiable generalization of Fisher's principle: the equilibrating quantity is the sex ratio at last birth rather than the sex ratio at birth. The algebraic identity in Eq. (4) is elegant, and the data collapse in Fig. 1c is a strong numerical result. The year-by-year simulations and the parameter bounds in Section V give the model some empirical contact, and the explicit prediction of a 1:1 SRLB is testable in historical or animal populations. However, the central stability claim is currently conditional on a fitness measure C^B that is introduced for this purpose, and the authors themselves show (Fig. 3) that p*_s is not an ESS under the standard whole-population gene-frequency measure at the second generation. The significance of the paper therefore depends on whether the C^B-based proof can be converted into a rigorous, standard invasion argument or clearly justified as the correct fitness currency for sex-ratio traits; at present the manuscript does not provide such a justification.","major_comments":[{"comment":"","section":"Section IV, Eq. (7) and Fig. 3"},{"comment":"","section":"Appendix B"},{"comment":"","section":"Section V, Eqs. (8)-(9)"}],"minor_comments":[{"comment":"","section":"Section III, Eq. (3)"},{"comment":"","section":"Section IV, Eq. (5)"},{"comment":"","section":"Section II"},{"comment":"","section":"Abstract and Section III"},{"comment":"","section":"Section II and Ref. [25]"},{"comment":"","section":"Section V"}],"recommendation":"major_revision","confidential_remarks":"The paper contains a valuable algebraic result and strong simulation support for the steady-state formula, but the ESS claim rests on a nonstandard fitness measure whose biological justification is not yet convincing, and the formal proof is a Mathematica black box with a placeholder code repository. I recommend major revision rather than rejection: the central claim is defensible in principle, but the manuscript needs a transparent derivation and a clear argument for why C^B is the correct invasion fitness, or a reformulation of the stability claim that does not rely on an unconventional measure. The authors should also be asked to provide the missing code and to recalibrate the human-application narrative to avoid presenting fitted parameters as predictions."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThe paper's core claim is worth your attention: when parents' probability of having another child depends on the last child's sex, the evolutionary equilibrium is not necessarily a balanced birth sex ratio, but a balanced sex ratio among last-born children. That is a new twist on Fisher, and it is testable. The steady-state formula p*_s = (1-b_d)/(2-b_s-b_d) is supported by a clean data collapse across 100 parameter combinations, and the last-child identity p*_s(1-b_s) = p*_d(1-b_d) is simple and likely correct. The simulations confirm the SRLB converges to 1:1. That part is solid.\n\nThe soft spot is the ESS proof. The paper introduces a matched-pair fitness measure C^B, the gene frequency among paired individuals, and then proves p*_s is an ESS under that measure. But the standard whole-population gene frequency X^B does not make p*_s an ESS at the second generation—the paper's own Fig. 3 shows that. The authors argue that at much later generations X^B peaks at Delta=0, but that is only a numerical observation, and the choice of generation is arbitrary. The Mathematica proof in Appendix B is a black box: no code released, and it only shows C^B is maximized, not that maximizing C^B is equivalent to uninvadability under standard population genetics. The stress-test note is right about this.\n\nThe human application inherits the fragility. They fit b_s and b_d to the observed 105:100 SRB, then the year-by-year simulation gives only 103:100, and they speculate about missing mortality details. That is a curve fit with assumptions, not a resolution of the puzzle. But the central last-born invariant does not collapse: it follows algebraically from the steady-state condition, independent of the ESS claim.\n\nThis is a paper for evolutionary biologists and demographers. It deserves serious refereeing—the model is clear, the prediction is falsifiable, and the data collapse is real. The referee should push for a genuine justification of C^B, or a reframing of the claim as a steady-state result rather than an ESS under standard invasion fitness. I would not cite it in my own work until the fitness measure is settled, but I would bring it to a reading group for the discussion.\n\nRecommendation: send it out, with major revision expected.\n\nBest,","headline":"A thoughtful extension of Fisher with a testable last-born sex-ratio prediction, but the ESS claim rests on a non-standard fitness measure that needs real justification.","tokens_in":12685,"tokens_out":3968,"would_cite":false,"duration_ms":43200,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["91A22","92D15"],"pacs":[],"model":"deepseek-v4-flash","headline":"Sex-ratio evolution is stabilized at the last-born child: the evolutionarily stable son probability is $p_s^* = (1-b_d)/(2-b_s-b_d)$, which makes last-born sons and daughters equally likely even when births are male-biased.","keywords":["sex ratio at birth","sex ratio among last-born children","evolutionarily stable strategy","Fisher's principle","son probability","continuation probabilities","human sex ratio","child mortality"],"falsifier":"Measure the sex of the last child in completed families in a historical or pre-modern population with a known male-biased birth sex ratio; the model predicts the last-born sex ratio will be approximately 1:1 whenever continuation probabilities differ after sons and daughters. If last-born sons are systematically overrepresented or underrepresented in such a population, the balancing identity $p_s^*(1-b_s)=p_d^*(1-b_d)$ would be contradicted as a description of evolved stopping rules.","tokens_in":11317,"feed_emoji":"👶","tokens_out":7535,"duration_ms":74109,"temperature":0.7,"pith_summary":"The paper asks why human births are persistently male-biased (about 105 boys per 100 girls) and proposes that the answer lies not in a 1:1 sex ratio at birth but in a 1:1 sex ratio among last-born children. In a game-theoretical model where a couple's probability of having another child depends on the sex of the previous child, the evolutionarily stable son probability is $p_s^* = (1-b_d)/(2-b_s-b_d)$, which is exactly the condition that the last child is equally likely to be a son or a daughter. Thus a biased birth sex ratio is the expected steady state whenever continuation probabilities differ after sons and daughters, while the sex ratio among last-born children stays balanced. The paper proves this balanced last-born ratio is an evolutionarily stable strategy under a fitness measure based on gene frequency within reproductively paired individuals, and it shows that sex-specific child mortality plus parents' tendency to try again after a death can place the human sex ratio at birth inside the model's predicted range.","feed_headline":"Last-born children balance sex ratio even when births run male","feed_subtitle":"ESS proof holds the last child's sex at 1:1, explaining why human births stay male-biased.","key_machinery":"The load-bearing identity is $p_s^*(1-b_s) = p_d^*(1-b_d)$: balancing the probability that a son is the last child against the probability that a daughter is the last child. Algebraically this gives $p_s^* = (1-b_d)/(2-b_s-b_d)$, so the model's equilibrium son probability is determined entirely by the two continuation probabilities. The stability proof is carried by the fitness measure $C^B$, the frequency of a gene among reproductively paired individuals (Eq. 7), rather than the whole-population gene frequency $X^B$; under $C^B$, a mutant son probability $p_s^* + \\Delta$ has lower fitness from the first generation onward, which is what establishes the evolutionarily stable strategy.","core_discovery":"The central discovery is a generalization of Fisher's principle: when the stopping rule for reproduction is sex-dependent, natural selection does not balance the sex ratio at birth; it balances the sex ratio among last-born children. Using a game-theoretical model with son probability $p_s$ and continuation probabilities $b_s$ (after a son) and $b_d$ (after a daughter), the paper derives the evolutionarily stable son probability $p_s^* = (1-b_d)/(2-b_s-b_d)$, which is equivalent to $p_s^*(1-b_s) = p_d^*(1-b_d)$, i.e., the probability that the last child is a son equals the probability that it is a daughter. Under the standard whole-population fitness measure, this value is not an evolutionarily stable strategy by the second generation, so the paper introduces the matched-pair gene frequency $C^B$ and proves that $p_s^*$ is a unique and globally stable evolutionarily stable strategy under this measure. The same mechanism, with higher male child mortality and higher continuation after a child's death, yields a male-biased birth ratio that brackets the observed human value of about 105 boys per 100 girls.","pith_inferences":["The paper's proof depends on choosing $C^B$ over the traditional whole-population fitness; if one regards the second-generation descendant count as the proper target of selection, the mathematical ESS result does not follow, even though the algebraic identity and simulations remain intact.","If empirical data on completed sibships show a last-born sex ratio at 1:1 despite a biased birth ratio, sex-ratio evolution would be better described as selection for balanced last births, with biased births an incidental consequence of mortality and continuation asymmetries.","A natural extension would let $b_s$ and $b_d$ themselves evolve alongside $p_s$; in that setting the balancing target could shift, for instance under cultural son preference or sex-selective stopping.","The prediction could be tested directly with modern birth-register data that record birth order and child survival: among families with completed reproduction, the last child's sex should be unbiased even in populations with an overall male-biased sex ratio at birth."],"forward_implications":["When $b_s = b_d$, the formula reduces to $p_s^* = 1/2$, so Fisher's classical 1:1 birth sex ratio is recovered as a special case.","If parents are more likely to continue after a son than after a daughter, the steady state has a male-biased sex ratio at birth while the sex ratio among last-born children remains 1:1.","The human sex ratio of about 105 boys per 100 girls falls inside the model's predicted range when male and female survival to adulthood are set near 0.5 and 0.55 and parents are more likely to have another child after a child's death.","The model predicts that any population with a biased birth sex ratio produced by sex-dependent continuation will still show an approximately balanced sex ratio among last-born children, a prediction the paper suggests testing on historical human populations or beef cattle.","Under the matched-pair fitness measure, $p_s^*$ is globally stable: no nearby alternative son probability can invade, so the balanced last-born ratio is a long-run attractor rather than a transient outcome."],"supporting_citations":[{"why":"Supplies the classical 1:1 sex-ratio principle that the paper generalizes.","marker":"[1]"},{"why":"Empirical estimate of the human sex ratio at birth that the model targets.","marker":"[12]"},{"why":"Cross-national estimates of male-biased human sex ratio at birth used as observed data for calibration.","marker":"[13]"},{"why":"Historical child-mortality estimate setting the survival probabilities $S_s=0.5$ and $S_d=0.55$.","marker":"[15]"},{"why":"Documents higher male than female childhood mortality, the biological basis for $b_s > b_d$.","marker":"[16]"},{"why":"Global mortality data supporting the assumed sex difference in infant survival.","marker":"[17]"},{"why":"Recent male versus female neonatal mortality data used when matching the model to the observed 105:100 sex ratio at birth.","marker":"[20]"}],"fun_headline_variants":["Last-born sex ratio hits 1:1 even when all births run male","Game-theory proof: evolution balances last child, not birth ratio","Child mortality and stopping rules explain male-biased human births","Fisher principle upgraded: last child’s sex ratio is the true target","Sex ratio puzzle solved: last-born offspring are always 50-50"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the new fitness measure $C^B$—gene frequency among reproductively paired individuals, rather than gene frequency in the whole population—is the correct measure of reproductive success for a sex-ratio trait; if that choice is rejected, the proof that the 1:1 last-born ratio is an evolutionarily stable strategy no longer goes through.","fun_headline_variants_meta":{"raw":{"variants":["Last-born sex ratio hits 1:1 even when all births run male","Game-theory proof: evolution balances last child, not birth ratio","Child mortality and stopping rules explain male-biased human births","Fisher principle upgraded: last child’s sex ratio is the true target","Sex ratio puzzle solved: last-born offspring are always 50-50"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000356,"raw_usage":{"total_tokens":1946,"prompt_tokens":973,"completion_tokens":973,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":589,"completion_tokens_details":{"reasoning_tokens":880}},"tokens_in":589,"tokens_out":973,"duration_ms":9700,"temperature":1.0,"reasoning_tokens":880,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T17:36:34.069816+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the sex of the last child in completed families in a historical or pre-modern population with a known male-biased birth sex ratio; the model predicts the last-born sex ratio will be approximately 1:1 whenever continuation probabilities differ after sons and daughters. If last-born sons are systematically overrepresented or underrepresented in such a population, the balancing identity $p_s^*(1-b_s)=p_d^*(1-b_d)$ would be contradicted as a description of evolved stopping rules.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the classical 1:1 sex-ratio principle that the paper generalizes."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Empirical estimate of the human sex ratio at birth that the model targets."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Cross-national estimates of male-biased human sex ratio at birth used as observed data for calibration."},{"cited_title":"Mortality in the past: every second child died,","cited_arxiv_id":null,"evidence_quote":"Historical child-mortality estimate setting the survival probabilities $S_s=0.5$ and $S_d=0.55$."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Documents higher male than female childhood mortality, the biological basis for $b_s > b_d$."},{"cited_title":"World population prospects 2024 – processed by our world in data,","cited_arxiv_id":null,"evidence_quote":"Global mortality data supporting the assumed sex difference in infant survival."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Recent male versus female neonatal mortality data used when matching the model to the observed 105:100 sex ratio at birth."}],"review_version":1}