{"id":"36dea769-9108-4074-9d01-0fd8105d5596","arxiv_id":"1908.06105","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":8,"one_line_summary":"The intrinsic scatter of the stellar radial acceleration relation is about 0.11 dex, in agreement with LCDM predictions and larger than the earlier null estimate.","lead":"Using over 2,500 spiral galaxies from six surveys, the authors measure how much of the observed spread in the radial acceleration relation comes from measurement error versus real cosmic variation. They find a real intrinsic scatter of about 0.11 dex, in line with dark matter galaxy formation models.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The inferred nonzero intrinsic scatter depends on an unvalidated Monte Carlo error model; a 1.7x increase in the adopted stellar M/L uncertainty alone removes it.","rationale":"The paper is a careful, good-faith effort: it assembles a large multi-survey sample, builds a Monte Carlo error model, and explicitly tests whether inflating individual error terms can erase the signal. The central inference is nevertheless fragile because the quadrature subtraction is only as good as the error model, and the error model's most influential term, the stellar M/L uncertainty, is also the least securely calibrated. Table 5's scaling-factor test is informative, but it varies one parameter at a time; a correlated underestimate across several parameters would require smaller individual inflations, and the paper does not report the global scaling factor needed for zero intrinsic scatter. The reader's CONDITIONAL verdict already captures this: the catalog and code are not released, and the error budget excludes several systematic contributions. My concern sharpens the reader's weakest assumption by identifying a specific, quantitative failure mode: a 1.7x increase in the M/L uncertainty, or a smaller correlated increase across several errors, brings the headline scatter to zero. This does not overturn the paper, but it is exactly the kind of sensitivity that should be demonstrated before the 0.11 dex value is treated as definitive.","tokens_in":20861,"tokens_out":7477,"duration_ms":87540,"concrete_test":"Rerun the full Monte Carlo pipeline with two modifications: (i) set the optical stellar mass-to-light ratio random uncertainty to 0.20 dex for all non-SPARC surveys, and (ii) apply a common multiplicative factor f to all parameter uncertainties (V, Υ, L, q, q0, D, m0) and find the smallest f for which the median inferred sigma_int is consistent with zero. If either change drives sigma_int below ~0.06 dex, the central claim of a robust 0.11 dex intrinsic scatter fails; if the median sigma_int remains above ~0.08 dex, the concern is resolved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 6.2 derives sigma_int = 0.11 +/- 0.02 dex by subtracting the Monte Carlo sigma_sim ~ 0.12 dex from the observed sigma_RAR ~ 0.17 dex in quadrature. The entire subtraction is controlled by the error model of Section 4, not by the data. Table 5 shows that enlarging the optical stellar mass-to-light ratio uncertainty by a factor of only 1.7 (from 0.13 to ~0.22 dex) makes the median intrinsic scatter zero; this is the smallest single-parameter scaling factor in the table. The adopted 0.13 dex random M/L uncertainty is taken from Roediger & Courteau (2015) as a representative upper value, but it is not validated for the heterogeneous PROBES galaxies, and the SHIVir/SPARC samples use different bands and M/L recipes. Literature estimates of the random (not systematic) scatter in stellar M/L reach ~0.2 dex. Moreover, the single-parameter stress test in Table 5 holds all other errors fixed; if the velocity, distance, and M/L errors are each modestly underestimated, their combined effect could remove the claimed intrinsic scatter with no individual factor as large as 1.7. The paper itself acknowledges (Sec. 6.2) that noncircular motions are not modeled and that some uncertainties may be underestimated; the beyond-one-Re check reduces but does not eliminate this sensitivity.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper develops a Monte Carlo model of observational errors to infer the intrinsic scatter of the stellar radial acceleration relation (RAR). Using a new PROBES compilation of roughly 2500 spiral galaxies from six surveys, the authors fit the MOND-inspired RAR (Eq. 1), project each galaxy onto the zero-scatter relation with a likelihood that accounts for shared parameters (Eq. 10), resample mock observations from the uncertainty model of Sec. 4, and process the mocks through the same fits and quality cuts as the data. Comparing the observed scatter (Table 1, median 0.17 dex) with the simulated scatter (Table 3, median 0.12 dex) in quadrature yields a median intrinsic scatter of 0.11 +/- 0.02 dex (Sec. 6.2), which the authors claim agrees with LCDM predictions of 0.06-0.08 dex and contradicts the null scatter reported by L17. Sensitivity tests in Table 5 and a beyond-one-Re check are presented in support of the result.","tokens_in":21159,"tokens_out":6274,"duration_ms":62039,"significance":"If the inference is robust, the paper would resolve an important controversy: it would show that the apparent tightness of the RAR is not exactly zero intrinsic scatter, alleviating the tension with LCDM galaxy formation simulations and weakening a simple MOND-based requirement. The study's strengths are its large, heterogeneous sample; the explicit modeling of shared per-galaxy errors; and the use of identical processing code for real and mock data, which avoids several common circularity pitfalls. The sensitivity table (Table 5) and the beyond-one-Re analysis are commendable. However, the central number is only as good as the error model in Sec. 4, and several components of that model are either unvalidated for the heterogeneous samples or defined inconsistently. These issues currently prevent me from endorsing the quantitative claim as established.","major_comments":[{"comment":"The distance uncertainty model is dimensionally inconsistent. The text states sigma_D = max(0.15D, 300), with 300 km/s being a peculiar velocity dispersion, while Eq. (4) uses sigma_D/D. As written, the first term treats sigma_D as a distance and the second as a velocity; the intended expression must involve sigma_pec/H_0 or an explicit unit conversion. Since Table 5 lists a distance-error scaling factor of 2.1, this ambiguity directly affects the inferred sigma_sim and sigma_int and must be corrected before the error model is usable.","section":"§4.1, Eq. (4)"},{"comment":"The mock galaxies are treated as spherically symmetric, while the observed g* values are computed from a flattened disk density (Eq. 7). The paper justifies this by noting that g* depends linearly on luminosity and mass-to-light ratio, but the disk geometry changes the radial profile of g* and hence the covariance structure of the forward residuals entering sigma_sim. The internal consistency argument only holds if the spherical simplification is demonstrated not to bias the scatter; I recommend running the Monte Carlo for at least one survey with a disky potential, or otherwise quantifying the effect.","section":"§5.2 and §2.2"},{"comment":"The derived intrinsic scatter is highly sensitive to the adopted random stellar mass-to-light ratio uncertainty. Table 5 shows that scaling that uncertainty by a factor of 1.7 (from 0.13 to roughly 0.22 dex) is sufficient to make the median sigma_int zero. The adopted 0.13 dex is a representative upper value from Roediger & Courteau (2015) for optical bands, not a validated value for each heterogeneous PROBES sample, and random M/L scatter estimates in the literature can reach about 0.2 dex. The single-parameter stress test also leaves open the possibility that modest combined underestimates of velocity, distance, and M/L errors remove the signal; a multi-parameter stress test or a prior-weighted marginalization over M/L uncertainty is needed to support the claim that the nonzero scatter is robust.","section":"§4.5 and Table 5"},{"comment":"Noncircular motions are not included in the error model. The beyond-one-Re check gives 0.10 +/- 0.01 dex, but this only removes the inner disk, where noncircular motions are strongest; the outer disk can still host warps, bars, or spiral perturbations that contribute to the observed scatter. A concrete test using the SPARC Q=1 versus Q=2 flags, or a comparison of H-alpha and H I rotation curves, would bound the magnitude of this neglected term. Without such a test, part of sigma_RAR may be misattributed to intrinsic scatter.","section":"§6.2 and §4.3"},{"comment":"The reported sample-median value conceals strong survey-to-survey variation. SPARC alone gives sigma_int = 0.040 +/- 0.013 dex, and several mass-binned entries in Table 7 are negative (sigma_int approximately -0.02 to -0.05 dex). The surveys with adopted rather than measured uncertainties (M92, M96, and to some extent C97) drive the median upward. The authors should either restrict the headline claim to samples with per-point errors, or justify why the heterogeneous and adopted-error samples can be combined in a single quadrature subtraction.","section":"§6.2, Tables 4 and 7"}],"minor_comments":[{"comment":"The exclusion of the 12 Q=3 SPARC galaxies is mentioned in Sec. 3.1.5, but the general quality-cut description in Sec. 3.2 does not state whether Q=3 objects are removed from the RAR analysis; please make this explicit.","section":"§3.1.5"},{"comment":"The text says axis ratio uncertainties are computed from isophotal ellipticity variations beyond Re, but the exact estimator (e.g., standard deviation of the mean, median absolute deviation) is not specified; Fig. 2 shows the resulting distributions but not the recipe.","section":"§4.2"},{"comment":"The maximum-likelihood projection in Eq. (10) is a methodological novelty, but the manuscript does not describe how the optimization is initialized or how convergence is verified for galaxies with many points; a brief technical note would improve reproducibility.","section":"§5.1"},{"comment":"The entries reported as '>10' for total luminosity and intrinsic disk flattening have no upper bound; please specify the search procedure or the maximum scaling factor that was tested.","section":"Table 5"},{"comment":"The claim that the stellar and baryonic RAR scatters are comparable is supported in part by a private communication (A. Dutton 2019); please replace this with a published or otherwise publicly documented analysis.","section":"§1"}],"recommendation":"major_revision","confidential_remarks":"The paper is timely and likely to be widely cited, but the central quantitative claim rests on the completeness of an error model whose most fragile components are acknowledged inside the paper. I would ask the authors to address the spherical-versus-disk inconsistency, fix the distance-error units, and strengthen the sensitivity analysis by considering combined or correlated error underestimates. The SPARC-only result and the negative mass-binned values in Table 7 should be reconciled with the headline median value."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nRead the Stone & Courteau RAR scatter paper. The headline: they build a Monte Carlo error model that finally treats the RAR's correlated, shared galaxy-level errors properly, run 2,600+ galaxies from six surveys through it, and infer an intrinsic stellar RAR scatter of 0.11 ± 0.02 dex — nonzero, in line with ΛCDM's 0.06–0.08 dex, and in tension with L17's claimed null. That's a new result, and the method is the real contribution. Anyone separating intrinsic from observational scatter in a scaling relation should borrow the approach.\n\nCredit where due: the mock data pass through the same quality cuts and fitting code as the real data; the forward/inverse axis issue that partly manufactured L17's tight relation is exposed clearly (Appendix A); and Table 5 is an honest stress test — the paper reports the factors by which each uncertainty would have to inflate to erase σ_int, and the smallest is 1.7 on the optical M/L error. They also check beyond one Re (0.10 dex) and flag noncircular motions and possible underestimated errors themselves. The circularity burden is genuinely low: the simulated scatter is forward-modeled, not tuned to match.\n\nNow the soft spot, and it is proportional: everything hinges on the completeness of that error model. A 1.7× inflation of the adopted 0.13 dex optical M/L uncertainty — the upper end of the Roediger & Courteau range, not independently validated for this heterogeneous catalog — zeros out the signal in the paper's own Table 5. The test holds other errors fixed while scaling one; if distance, velocity, and M/L are each modestly underestimated together, no individual factor needs to reach 1.7. The quoted ±0.02 dex propagates only survey-to-survey scatter, not this systematic sensitivity. Also telling: the best-fit g† values in Table 1 differ by factors up to ten across surveys, which says the M/L modeling carries survey-level systematics much larger than the adopted random term. That does not directly bias a scatter measurement, but it should temper confidence in the error model's completeness. The PROBES catalog and code are not yet released — a real obstacle for a result that leans this heavily on an error model.\n\nMy take: the direction of the claim — nonzero intrinsic scatter around a tenth of a dex — is probably right and the paper deserves a serious referee; it was in fact published in ApJ. But the specific 0.11 ± 0.02 is softer than the abstract implies. If I were refereeing, I'd ask for the catalog release and a joint multi-parameter sensitivity analysis. I'd bring it to reading group and I'd cite the method.","headline":"A careful Monte Carlo error model on a 2,500-galaxy catalog yields a nonzero RAR intrinsic scatter around 0.11 dex that favors LCDM over L17's null, but the value leans heavily on the adopted M/L uncertainty and is less secure than the abstract's ±0.02 suggests.","tokens_in":21736,"tokens_out":5663,"would_cite":true,"duration_ms":53553,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The radial acceleration relation carries intrinsic scatter of 0.11 dex.","keywords":["radial acceleration relation","intrinsic scatter","Monte Carlo error model","galaxy scaling relations","galaxy formation","LCDM predictions","MOND","spiral galaxy rotation curves"],"falsifier":"Measure the stellar mass-to-light ratio independently per galaxy with spatially resolved stellar-population modeling on a PROBES subsample and include noncircular motions modeled from high-resolution simulations; if the simulated scatter from these errors alone rises to the observed $0.17$ dex, the intrinsic scatter is zero, and the $0.11$ dex claim collapses.","tokens_in":20613,"feed_emoji":"🔭","tokens_out":5330,"duration_ms":46515,"temperature":0.7,"pith_summary":"The paper claims that the radial acceleration relation (RAR)—the tight empirical link between a galaxy's observed rotation acceleration and the acceleration expected from its stars alone—has a real intrinsic scatter of $0.11\\pm0.02$ dex once measurement errors are removed. This matters because a previous analysis of the baryonic RAR reported near-zero intrinsic scatter, a result that favored modified Newtonian dynamics over the standard cosmological model. The authors assembled more than 2500 spiral galaxies from six surveys and modeled every major source of observational uncertainty with a Monte Carlo simulation. Subtracting that simulated error scatter in quadrature from the observed $0.17$ dex scatter leaves $0.11\\pm0.02$ dex, consistent with ΛCDM predictions of $0.06$–$0.08$ dex. If correct, the RAR's tightness is real but not perfect, and it no longer singles out modified gravity.","feed_headline":"Radial acceleration relation has real intrinsic scatter, 0.11 dex","feed_subtitle":"A careful error model leaves a cosmic scatter that galaxy formation can explain—not the perfectly tight relation MOND needs.","key_machinery":"The load-bearing machinery is a Monte Carlo error model that simulates a universe with zero intrinsic scatter and then asks how much scatter observational errors alone would produce. Each observed point is first projected onto the fitted RAR by maximizing a likelihood over all measured parameters—distance, inclination, intrinsic disk flattening, velocity, luminosity, and stellar mass-to-light ratio—while requiring the point to lie exactly on the relation. Those zero-scatter values are then resampled from their uncertainty distributions, and the mock data are run through the same quality cuts and fitting pipeline as the real data. Comparing the mock scatter to the observed scatter in quadrature isolates the intrinsic component, a step that a simple first-order error propagation cannot do because many uncertainties are shared across points within a galaxy.","core_discovery":"The central discovery is that the apparent tightness of the RAR hides a non-zero cosmic scatter whose size matches the expectations of cosmological galaxy formation. Using the forward residuals of the stellar RAR, the median observed scatter across six surveys is $0.17$ dex, while a Monte Carlo model that resamples all known observational errors produces a scatter of $0.12$ dex. The difference in quadrature gives an intrinsic scatter of $0.11\\pm0.02$ dex, close to, though slightly larger than, the $0.06$–$0.08$ dex predicted by ΛCDM simulations and incompatible with the null value previously inferred. The intrinsic scatter also decreases with galaxy mass, from about $0.14$ dex at low mass to about $0.10$ dex at high mass, mirroring simulation trends.","pith_inferences":["If the intrinsic scatter is genuinely near $0.1$ dex, any modified-gravity theory that demands a perfectly tight relation must be either revised or supplemented by an astrophysical source of scatter.","The same Monte Carlo projection-and-resample recipe could be applied to the baryonic RAR once gas maps are available for all PROBES galaxies; a similar estimate would directly test the assumption that stellar and baryonic RARs share their scatter.","The method's logic implies a falsifiable hierarchy: future higher-resolution, lower-error surveys should show the observed scatter shrinking toward the simulated error scatter, keeping the quadrature residual near $0.1$ dex rather than falling to zero.","The mass-dependent trend of the intrinsic scatter suggests that a single universal acceleration scale may be an oversimplification; the low-mass tail is where the relation's scatter is largest and where tests of universality should focus."],"forward_implications":["A non-zero intrinsic scatter of about $0.11$ dex replaces the null value reported for the baryonic RAR, so the relation no longer discriminates against ΛCDM galaxy formation.","To reduce the intrinsic scatter to zero, the measurement uncertainties would have to be inflated by factors of about 2 for the stellar mass-to-light ratio up to factors greater than 10 for other parameters.","The intrinsic scatter decreases with galaxy mass (median $0.144$, $0.103$, and $0.095$ dex from low to high mass), matching the mass dependence seen in ΛCDM simulations and the diversity of dwarf rotation curves.","In the outer regions beyond one effective radius, where noncircular motions are weaker, the intrinsic scatter is $0.10\\pm0.01$ dex, confirming the result.","The forward versus inverse choice of residuals changes the measured scatter substantially (median forward $0.17$ dex versus inverse $0.24$ dex), a consideration that affects all galaxy scaling relations."],"supporting_citations":[{"why":"Reports the null intrinsic scatter of the baryonic RAR from the SPARC sample that this paper re-examines and rejects.","marker":"Lelli et al. (2017)"},{"why":"Introduces the RAR and its one-parameter fitting form used throughout.","marker":"McGaugh et al. (2016)"},{"why":"Provides the SPARC data set and its reported uncertainties used as one of the six surveys.","marker":"Lelli et al. (2016)"},{"why":"Supplies the $0.09$–$0.13$ dex random uncertainty in optical stellar mass-to-light ratios adopted in the error model.","marker":"Roediger & Courteau (2015)"},{"why":"One of the ΛCDM simulations predicting $0.06$–$0.08$ dex intrinsic baryonic RAR scatter.","marker":"Keller & Wadsley (2017)"},{"why":"Another ΛCDM simulation prediction for the intrinsic baryonic RAR scatter that the measured value is compared with.","marker":"Ludlow et al. (2017)"},{"why":"Predicts mass-dependent baryonic RAR scatter and motivates the mass-binned analysis.","marker":"Dutton et al. (2019)"},{"why":"Supplies the $0.11$ dex uncertainty for the $3.6\\,\\mu$m stellar mass-to-light ratio used for SPARC galaxies.","marker":"Meidt et al. (2014)"},{"why":"Contributes the M96 survey of 1216 galaxies, a major share of the PROBES sample.","marker":"Mathewson & Ford (1996)"}],"fun_headline_variants":["RAR's hidden scatter: 0.11 dex, not zero, matches LCDM","Radial acceleration relation has real scatter, 0.11 dex","Intrinsic scatter in RAR found: 0.11 dex, supporting LCDM","RAR isn't perfectly tight: 0.11 dex scatter, MOND challenged"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The Monte Carlo error model has to capture every significant observational uncertainty—especially the $0.13$ dex stellar mass-to-light ratio scatter, inclination recovery from axis ratios, magnitude errors, and the neglect of noncircular motions—because any underestimate inflates the claimed intrinsic scatter.","fun_headline_variants_meta":{"raw":{"variants":["RAR's hidden scatter: 0.11 dex, not zero, matches LCDM","Radial acceleration relation has real scatter, 0.11 dex","Intrinsic scatter in RAR found: 0.11 dex, supporting LCDM","RAR isn't perfectly tight: 0.11 dex scatter, MOND challenged"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000316,"raw_usage":{"total_tokens":1821,"prompt_tokens":1006,"completion_tokens":815,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":622,"completion_tokens_details":{"reasoning_tokens":727}},"tokens_in":622,"tokens_out":815,"duration_ms":7338,"temperature":1.0,"reasoning_tokens":727,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:55:44.178441+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the stellar mass-to-light ratio independently per galaxy with spatially resolved stellar-population modeling on a PROBES subsample and include noncircular motions modeled from high-resolution simulations; if the simulated scatter from these errors alone rises to the observed $0.17$ dex, the intrinsic scatter is zero, and the $0.11$ dex claim collapses.","supporting_citations":[{"cited_title":"S., Lelli, F., & Schombert, J","cited_arxiv_id":null,"evidence_quote":"Introduces the RAR and its one-parameter fitting form used throughout."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the $0.09$–$0.13$ dex random uncertainty in optical stellar mass-to-light ratios adopted in the error model."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"One of the ΛCDM simulations predicting $0.06$–$0.08$ dex intrinsic baryonic RAR scatter."},{"cited_title":"D., Ben´ ıtez-Llambay, A., Schaller, M., et al","cited_arxiv_id":null,"evidence_quote":"Another ΛCDM simulation prediction for the intrinsic baryonic RAR scatter that the measured value is compared with."},{"cited_title":"A., Macci` o, A","cited_arxiv_id":null,"evidence_quote":"Predicts mass-dependent baryonic RAR scatter and motivates the mass-binned analysis."},{"cited_title":"E., Schinnerer, E., van de Ven, G., et al","cited_arxiv_id":null,"evidence_quote":"Supplies the $0.11$ dex uncertainty for the $3.6\\,\\mu$m stellar mass-to-light ratio used for SPARC galaxies."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Contributes the M96 survey of 1216 galaxies, a major share of the PROBES sample."}],"review_version":1}