{"id":"ba229186-ed3d-49c5-9363-93b72807e107","arxiv_id":"2509.02940","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"high","formal_verification":"none","parameter_count":5,"one_line_summary":"Cambrian fossil trails show autocorrelated turning angles, Ediacaran trails do not, which the authors interpret as early time-tuned behaviour.","lead":"Researchers measured whether turning angles in fossil animal trails repeat over time by applying autocorrelation to Ediacaran and Cambrian grazing paths. They found autocorrelation in two Cambrian ichnospecies but not two Ediacaran ones, and interpret this as the earliest sign of time-tuned, possibly memory-based behaviour.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The distance-to-time mapping rests on an unverified constant-velocity assumption; without it, spatial autocorrelation cannot be read as temporal memory.","rationale":"The reader's weakest assumption correctly identifies the constant-average-velocity premise as load-bearing. My independent review converges on the same point: the paper's leap from 'autocorrelation in turning angles along equidistant spatial segments' to 'temporal correlation / time-tuned behaviours' is only valid if each segment represents a constant unit of time. The paper explicitly states this as an assumption but provides no validation. Other concerns—lack of null-model controls, use of Pearson correlation on circular angles, small sample size—are real, but they would matter less if the temporal interpretation were secure; they would affect the statistical strength, not the conceptual mapping. Conversely, if the constant-velocity assumption is false, the central claim is not merely underpowered but conceptually invalid: distance-based autocorrelation cannot speak to temporal memory. Thus this is the single most load-bearing concern. A concrete empirical test using extant movement data can directly assess whether the distance-based discretization is a reliable proxy for temporal autocorrelation, making the resolution of this concern tractable. Given that the assumption is unverified and central, the REJECT verdict stands.","tokens_in":18948,"tokens_out":3557,"duration_ms":46088,"concrete_test":"Use high-resolution movement data from a modern benthic grazer (or an agent-based simulation with known time stamps, variable speed, and turning behaviour) to test the constant-velocity mapping. From the true time-stamped path, compute the turning-angle autocorrelation in two ways: (1) using the paper's method—resample at equal distances based on the overall average velocity, then compute autocorrelation over distance lags; and (2) using the true equal-time sampling, computing autocorrelation over time lags. Compare the resulting autocorrelation functions (sign, magnitude, and lag structure). If the distance-based proxy fails to reproduce the true time-based autocorrelation (e.g., shows spurious positive or negative autocorrelation where the time-based function is flat), the constant-velocity assumption is invalid, and the fossil inferences about temporal memory are unsupported. If the t","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—that Ediacaran trails show no temporal autocorrelation while Cambrian trails do, indicating time-tuned behaviours—depends entirely on converting spatial lag distances into temporal lags. The Methods state: 'we assumed that an average velocity provided a reasonable approximation of the velocity distributions of each tracemaker,' and then 'Each equidistant segment can be interpreted to represent the passage of an approximately constant unit of time.' This is the load-bearing step. If the tracemaker's speed varied during trail formation (pausing, slowing at turns, speeding on straights, or responding to substrate), then equal spatial segments correspond to unequal time intervals. The autocorrelation of turning angles computed from distance-based lags would then mix spatial and temporal structure; a pattern the paper interprets as 'temporal autocorrelation' and 'memory' could be an artifact of speed variation and path geometry. The paper offers no independent evidence for constant velocity, and trace fossils record only the path, not the pace. The authors do acknowledge that autocorrelation can arise from other factors (Introduction), but their defence—that only strongly periodic external cues or single-stroke gaits could cause it—does not address the possibility that variable speed, without any temporal memory, generates distance-based autocorrelation. Because the entire temporal interpretation collapses if this assumption fails, this is the most load-bearing concern.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript applies autocorrelation and partial autocorrelation analyses to turning angles extracted from four sets of fossil grazing trails: two Ediacaran ichnospecies (Helminthoidichnites tenuis, Parapsammichnites pretzeliformis) and two early Cambrian ichnospecies (Psammichnites cf. saltensis, Psammichnites gigas circularis). Trails are digitized, divided into equal-length segments scaled by trail width, and the authors assume that each segment corresponds to an approximately constant unit of time. They report that the Ediacaran trajectories show no temporal autocorrelation beyond short lags, whereas both Cambrian trajectories show positive and/or negative autocorrelation over longer lags. This is interpreted as evidence that time-tuned behaviours and possible biological memory appeared by the early Cambrian, supporting the Cambrian Information Revolution hypothesis. The paper also proposes that the length of sampling-induced positive autocorrelation can be used to infer an organism's 'action step'.","tokens_in":19321,"tokens_out":4204,"duration_ms":51368,"significance":"If the central inference were valid, this would be a valuable quantitative contribution to debates about behavioural and cognitive evolution across the Ediacaran–Cambrian transition. The study is conceptually novel in applying autocorrelation methods, common in movement ecology, to fossil trails, and it makes explicit, falsifiable predictions about Ediacaran versus Cambrian behaviour. The authors are also transparent about some interpretative caveats, such as the distinction between statistical and biological memory. However, the central claim that the observed spatial autocorrelation reflects temporal, time-tuned behaviour depends on an unverified and likely unverifiable assumption of constant movement speed, and the analysis lacks formal hypothesis tests and null-model controls. These issues are load-bearing: without them, the paper documents spatial autocorrelation patterns in trace fossils, but not the temporal memory or time-tuned navigation claimed in the title and abstract.","major_comments":[{"comment":"The distance-to-time mapping is the core assumption of the paper, and it is unverified. The authors state that they 'assumed that an average velocity provided a reasonable approximation of the velocity distributions of each tracemaker' and that 'Each equidistant segment can be interpreted to represent the passage of an approximately constant unit of time.' Trace fossils preserve only the path, not the pace. If the tracemaker varied its speed while forming the trail—pausing, slowing at turns, or reacting to substrate—equal spatial segments would correspond to unequal time intervals. The autocorrelation of turning angles computed from distance-based lags would then mix spatial and temporal structure, and the patterns interpreted as 'temporal autocorrelation' and 'time-tuned behaviour' could arise from speed variation without any memory or timing mechanism. No sensitivity analysis, simulati","section":"§4.3 Methods, 'For our specimens we assumed...'"},{"comment":"The 'action step' is inferred from the very autocorrelation curves that are subsequently interpreted. The text states that the distance at which sampling-induced autocorrelation ends 'could indicate the typical distance travelled in a single action by the organism,' and the turning angles are then re-calculated and re-analysed using these inferred action steps. This is circular: the unit used to define action-step lags is derived from the same data that are then used to claim autocorrelation at certain action-step lags. The reported 'similar trend' in the action-step re-analysis is therefore expected by construction and does not independently validate the existence of action-scale temporal memory. An independent definition of an action step, or a demonstration that the results are robust to alternative choices of this parameter, is needed.","section":"§4.4 Results and Figure 4.5"},{"comment":"The manuscript lacks formal hypothesis tests, confidence intervals, and null-model controls. The analysis reports mean autocorrelation coefficients, standard errors of the mean, and boxplot summaries, and then makes claims such as 'greater evidence in favour of significant anticorrelation' (§4.4) without any significance test. A small standard error of the mean across specimens does not test whether the mean autocorrelation differs from zero or from the expected value under a null model. No null model is presented for the autocorrelation function of a random walk with the same path length, discretization, and measurement protocol. Without such a null, the observed Cambrian–Ediacaran difference cannot be distinguished from artifacts of segment length, trail geometry, or sample size. The number of specimens per ichnospecies is also not stated in the Methods or Results, so the statistical p","section":"§4.4 Results and §4.5 Discussion"},{"comment":"The paper's interpretive axiom—that autocorrelation at time lags larger than the action step is internally driven and evidences time-tuned behaviour or biological memory—is asserted rather than established. The authors dismiss non-periodic external factors such as nutrient distribution or sediment consistency as unlikely to produce 'mathematically repeating locomotory patterns,' but the analysis actually measures spatial autocorrelation of turning angles, not periodicity. A meandering or looping trail produced by local taxis, by following a chemical gradient, or by a simple mechanical interaction with the substrate can generate positive and negative autocorrelations at multiple lags without any internal timer or memory. The paper's own discussion acknowledges that external memory systems (e.g., slime mould markers) can produce complex trajectories, yet it does not explain why such mechan","section":"§4.2 Introduction, paragraph beginning 'The presence of temporal autocorrelation...'"}],"minor_comments":[{"comment":"There is a typo: 'Parapsammichnits pretzleformis' should be 'Parapsammichnites pretzeliformis'.","section":"§4.3 Methods, ichnospecies list"},{"comment":"The caption lists two panels labelled '(B)': one for the trajectory and one for the scatterplots. The second should presumably be '(D)' or another letter, and the panel labels in the figure should be checked.","section":"Figure 4.2 caption"},{"comment":"The text refers to 'Supplementary Information, Figures 4.10 & 4.11' and 'Figures 4.11 & 4.12' for partial autocorrelation functions, but the supplementary figures are numbered 4.16–4.19. These cross-references should be corrected.","section":"Supplementary Information references"},{"comment":"The discretization method is repeatedly referenced as 'Chapter 3' (a thesis chapter) rather than described in sufficient detail. Since this method is central, the key algorithmic steps, including image-to-curve conversion and segment resampling, should be summarized in the paper or appendices, and the referenced chapter should be made available.","section":"§4.3 Methods, 'Chapter 3'"},{"comment":"The text states that mean r and standard error data are available in Supplementary Data, but no data table or repository link is included in the manuscript. Given the quantitative nature of the claims, the raw autocorrelation values, specimen counts, and code should be provided in a permanent repository.","section":"Data availability"}],"recommendation":"reject","confidential_remarks":"The manuscript would be improved if it were reframed as a study of spatial autocorrelation in fossil trails, with the temporal and memory-related claims clearly separated from the spatial measurements. As written, the central inference rests on an unverifiable constant-velocity assumption and a circular action-step definition, and the lack of null-model controls makes the Cambrian–Ediacaran difference difficult to evaluate. These are load-bearing issues that the current scope of the paper cannot fix without substantial additional evidence or a fundamental reframing."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know this paper does something genuinely new: it applies autocorrelation and partial autocorrelation to fossil grazing trails, extracting quantitative temporal structure from trace fossils. The raw pattern—two Cambrian ichnospecies showing autocorrelation in turning angles while two Ediacaran ones do not—is the kind of descriptive data the field needs more of. The authors are also honest about some limits. They explicitly distinguish mathematical memory from biological memory, and they acknowledge autocorrelation could come from mechanics or external periodicities. That is more self-awareness than many paleo papers manage.\n\nThe problem is the step from spatial segments to time. The method discretizes trails into equal-distance segments and assumes each segment represents a constant unit of time, based on an unverified average-velocity assumption. The stress-test note is right: if the tracemaker paused, slowed at turns, sped up on straights, or responded to substrate, equal distance does not mean equal time. Trace fossils preserve the path, not the pace. Without independent evidence for constant velocity, the autocorrelation could be a spatial artifact, not a temporal one. The paper acknowledges other causes but never addresses variable speed. This is a load-bearing gap.\n\nThere are other soft spots. The core discretization method is in an unpublished 'Chapter 3', so the procedure is not independently checkable. Turning angles are circular data, but the authors use linear Pearson correlation, which can distort relationships near the ±π boundary. The design is two ichnospecies per era, which is thin for a Cambrian Information Revolution claim, and there are no null-model controls or formal hypothesis tests. They also fit the 'action step' from the autocorrelation curves and then use it as the unit for subsequent analysis—that is somewhat circular, though they are transparent about it.\n\nNone of this means the raw measurements are fake. The technique is promising, and the paper is a reasonable first pass at quantifying behavioral memory in ichnology. But the headline claim—'time-tuned behaviours were in place by the early Cambrian'—is not supported as presented. It needs a null model with variable speed, formal tests against simulated paths, or independent calibration of the distance-to-time link.\n\nI would send this to peer review, because the idea is fresh and the data are real, but I would expect major revision. For a reading group, it is worth discussing as a cautionary example of spatial-temporal inference. I would not cite it as evidence for memory in early animals.","headline":"Clever quantitative idea applied to fossil trails, but the distance-to-time mapping is load-bearing and unverified, so the memory conclusion overreaches.","tokens_in":19758,"tokens_out":2127,"would_cite":false,"duration_ms":25806,"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":"Fossil trail turns show early Cambrian grazers already used time-tuned, memory-like behaviour while Ediacaran trails show no temporal correlation.","keywords":["Ediacaran-Cambrian transition","trace fossils","autocorrelation","turning angle","movement ecology","time-tuned behaviour","Cambrian Information Revolution","sensory evolution"],"falsifier":"Track a living grazing or burrowing animal whose speed is known to vary, discretize its path under the same constant-average-velocity assumption, and compute the autocorrelation of turning angles. If a mechanically generated or memory-free path yields positive-then-negative autocorrelation like the Cambrian trails, the inference collapses. A complementary check is to simulate paths with and without time-tuned turning rules and ask whether the method reliably separates them at fossil sample sizes.","tokens_in":18880,"feed_emoji":"🐾","tokens_out":10597,"duration_ms":112785,"temperature":0.7,"pith_summary":"The paper claims that the pattern of turns in fossil grazing trails can reveal when ancient animals began to use time-tuned, memory-like behaviour, and that this happened by the early Cambrian. It analyses four sets of trails: two Ediacaran trace-fossil species, whose turning angles show no meaningful correlation across time lags, and two early Cambrian trace-fossil species, whose turning angles are correlated over several trail-widths. The central result is a contrast: no temporal autocorrelation in the Ediacaran trails, clear autocorrelation in the Cambrian ones. If the interpretation holds, it supports the Cambrian Information Revolution hypothesis—that an increasingly information-rich early Cambrian world selected for new cognitive and behavioural strategies, not just new body plans—and gives palaeontologists a quantitative way to study the evolution of navigation.","feed_headline":"Trail turns reveal early Cambrian grazers used time-tuned behaviour","feed_subtitle":"Ediacaran trails show no temporal correlation; Cambrian trails do—memory-like navigation before brains and eyes.","key_machinery":"The key object is the autocorrelation function of turning angles along a digitized fossil path. Trails are traced and divided into segments whose length is a multiple of the trail width; under a constant average velocity, each segment is treated as roughly one unit of time. The autocorrelation at lag h asks whether a turn predicts a turn h units later, and partial autocorrelation removes the influence of the lags in between. The span of an initial sampling-induced positive correlation is used to estimate the distance of one 'action step'—a single locomotory cycle—and any correlation that persists beyond that span is read as evidence for internally driven, time-tuned behaviour.","core_discovery":"The central discovery is a statistical signature in fossil movement paths. After discretizing trails into equal segments and computing turning angles, the autocorrelation function shows that in Helminthoidichnites tenuis and Parapsammichnites pretzeliformis from the Ediacaran, turns are essentially independent of turns taken one to several steps earlier, apart from a brief sampling-induced positive correlation and a tentative short-range anticorrelation. In Psammichnites cf. saltensis and Psammichnites gigas circularis from the early Cambrian, turns are positively autocorrelated at short lags, and P. cf. saltensis shows a prolonged anticorrelation extending out to roughly 28 trail-widths. Th","pith_inferences":["Beyond the paper: applying the same analysis to later Phanerozoic trails could map when time-tuned behaviours spread across environments and lineages.","Beyond the paper: modern tracking data on variable-speed grazers could test whether the constant-velocity assumption is safe; if it is not, the spatial-to-temporal mapping would need a velocity model.","Beyond the paper: coupling autocorrelation with periodicity detection could separate clock-driven rhythms from memory-like search strategies, a distinction the paper notes is unresolved."],"forward_implications":["Autocorrelation analysis of fossil trails can serve as a new proxy for navigation capacity, extending the record of sensory evolution beyond body fossils and qualitative trail descriptions.","Similar-looking trails can hide different temporal strategies: the two 'looping' ichnospecies, one Ediacaran and one Cambrian, have different autocorrelation patterns, so morphological similarity alone is not a reliable guide to behaviour.","The decay length of the sampling-induced autocorrelation gives an estimate of the action-step distance, potentially a general way to infer locomotory cycle lengths from fossil trails.","Time-tuned behaviour in the early Cambrian would support the idea that increasing environmental information complexity—not only ecological or anatomical change—was a distinct evolutionary force in the Cambrian radiation."],"supporting_citations":[{"why":"Supplies the Cambrian Information Revolution hypothesis that frames the paper's predictions.","marker":"Plotnick et al., 2010"},{"why":"Establishes the classic inference that regular Psammichnites trails reflect programmed behaviour, which the paper tests quantitatively.","marker":"Seilacher, 1967"},{"why":"Provides the Movement Ecology Paradigm linking trail geometry to navigation capacity, motion capacity, internal state, and external factors.","marker":"Nathan et al., 2008"},{"why":"Documents autocorrelation methods in animal-movement studies and the sampling-induced correlation used to define action steps.","marker":"Dray et al., 2010"},{"why":"Describes Parapsammichnites pretzeliformis as a terminal-Ediacaran bulldozer able to sense nutrient-rich regions, anchoring the Ediacaran interpretation.","marker":"Buatois et al., 2018"},{"why":"Provides prior qualitative trace-fossil evidence of increasing behavioural complexity through the Ediacaran–Cambrian that this paper quantifies.","marker":"Carbone & Narbonne, 2014"},{"why":"Distinguishes biological memory from statistical memory and lays out when spatial memory should be adaptive.","marker":"Fagan et al., 2013"},{"why":"Documents the absence of complex sensory organs in Ediacaran fossils, framing the evolutionary gap the results help fill.","marker":"Evans et al., 2021"},{"why":"Shows that externalized spatial memory in a non-neuronal organism can support navigation, making cheap time-tuned behaviour plausible.","marker":"Reid et al., 2012"}],"fun_headline_variants":["Trail turns reveal Cambrian grazers had time-tuned behavior","Ediacaran tracks lack temporal memory; Cambrian show it","Fossil paths show early Cambrian animals used temporal sense","Cambrian trails reveal memory-like navigation, Ediacaran don't","Turning angles: Cambrian grazers behaved with time in mind"],"cache_read_input_tokens":2688,"weakest_assumption_plain":"The load-bearing premise is that distance along the trail is a clock: the analysis assumes each tracemaker moved at roughly constant average speed, so equal distances stand for equal time intervals, and it further assumes that any correlation persisting beyond one locomotory cycle is an internally generated behavioural signal rather than a mechanical, geometric, or substrate artefact.","fun_headline_variants_meta":{"raw":{"variants":["Trail turns reveal Cambrian grazers had time-tuned behavior","Ediacaran tracks lack temporal memory; Cambrian show it","Fossil paths show early Cambrian animals used temporal sense","Cambrian trails reveal memory-like navigation, Ediacaran don't","Turning angles: Cambrian grazers behaved with time in mind"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000238,"raw_usage":{"total_tokens":1327,"prompt_tokens":703,"completion_tokens":624,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":447,"completion_tokens_details":{"reasoning_tokens":536}},"tokens_in":447,"tokens_out":624,"duration_ms":7493,"temperature":1.0,"reasoning_tokens":536,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T11:15:13.464133+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Track a living grazing or burrowing animal whose speed is known to vary, discretize its path under the same constant-average-velocity assumption, and compute the autocorrelation of turning angles. If a mechanically generated or memory-free path yields positive-then-negative autocorrelation like the Cambrian trails, the inference collapses. A complementary check is to simulate paths with and without time-tuned turning rules and ask whether the method reliably separates them at fossil sample sizes.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides prior qualitative trace-fossil evidence of increasing behavioural complexity through the Ediacaran–Cambrian that this paper quantifies."}],"review_version":1}