{"id":"c58c2964-846f-4d18-b777-ac224a64d3a3","arxiv_id":"1908.05736","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A design principle called SURFPAD can create profoundly ambiguous color displays at will, and the perception of the new crocs and socks images is predicted by fabric and lighting priors.","lead":"This paper introduces a design principle for creating color images, such as photos of crocs and socks, that different people reliably see in different colors. It shows that what people see depends on assumptions, such as whether the socks look white, which are shaped by experience.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The white-sock prior's causal role in P2 is not established: the sock report followed the P2 judgment, allowing reverse causation, and no counterbalanced order or IMO control was run.","rationale":"The paper's demonstration that five SURFPAD-created croc-and-sock images are categorically ambiguous, while the same objects under typical light yield ~98% consensus, is well supported by the large online sample and is not the fragile part of the manuscript. The fragile part is the explanatory layer: the abstract claims that 'we can predict the percept from fabric priors resulting from experience.' The evidence for this is the correlation between sock-white reports and P2-pink reports (Fig. 7) and the weaker correlation between self-reported tube-sock wearing and sock-white reports (Fig. 8). The reader correctly identifies the causal-priority problem: the sock question was asked immediately after the P2 judgment, so the P2 percept could anchor the sock report. A consistency effect — 'I saw pink, so the socks must be white' — would produce exactly the observed 12-fold association without any pre-existing fabric prior. The paper acknowledges the order issue only for subsequent stimuli (the sock question 'might have drawn attention to the color of the socks'), not for the reverse path from P2 to the sock report. The Discussion concedes the missing no-sock condition. This makes the mechanistic claim CONDITIONAL, which is exactly the reader's verdict. My independent read converges on the same load-bearing assumption; therefore I recommend no change to the verdict.","tokens_in":14770,"tokens_out":11008,"duration_ms":103681,"concrete_test":"Run the survey with two between-subjects orders: (A) P2 croc judgment followed by the sock-white question (the current order), and (B) the sock-white question before the P2 judgment, ideally before any croc display is shown. If the association between 'socks seen as white' and 'P2 seen as pink' is substantially attenuated or absent in order B, the claim that a pre-existing white-sock prior drives the P2 percept fails; if the association persists, reverse causation from P2 to the sock report is ruled out.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central inference that a pre-existing white-sock fabric prior drives the P2 croc percept is not secured by the reported data. The only supporting evidence is a cross-sectional correlation between the sock-white report and the P2-pink report (Results, 'What determines the perception of the crocs displays?', Fig. 7), collected in a fixed order in which the sock question immediately follows the P2 judgment. With this order, the association could reflect the P2 percept anchoring the subsequent sock report (reverse causation) or a shared third factor such as a general response bias; the paper provides no counterbalanced order, no no-sock control, and no manipulation of the IMO to break this symmetry. The Discussion concedes the missing no-sock condition ('it stands to reason that we should create and ask observers about crocs altogether without socks'), and the sock-wearing self-report link (Fig. 8) is admitted to be 'obviously less strong than what would be needed to comfortably claim that we understand this aspect.' Thus the headline claim 'predict the percept from fabric priors resulting from experience' rests on a single-order correlation, not on evidence that the prior is causally prior.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a design principle, SURFPAD (Substantial Uncertainty combined with Ramified or Forked Priors and Assumptions yields Disagreement), for creating color displays that evoke categorical disagreement among observers. The authors created five crocs-and-socks images under altered lighting (P0, P1, P2, M1, M2) and surveyed 5,762 online participants. They report that the same crocs under typical lighting yield about 98% consensus, whereas the five altered images produce categorically different modal percepts across observers. The paper then examines correlates of P2 perception, reporting that participants who saw the socks as white were much more likely to see P2 as pink (Fig. 7), that self-reported tube-sock wearing is weakly associated with sock and P2 perception (Fig. 8), and that P2 perception strongly predicts sneaker perception but not dress perception (Fig. 9). The authors conclude that differential fabric priors, not illumination priors, drive the crocs disagreement and suggest that such priors underlie polarized disagreement more broadly.","tokens_in":14943,"tokens_out":3892,"duration_ms":41323,"significance":"If the design principle holds, the paper makes a useful methodological contribution: previous color-ambiguous displays such as the dress and the sneaker arose spontaneously, whereas this paper demonstrates a repeatable procedure for creating such displays. The large sample, bootstrap confidence intervals, and the typical-light control condition (98% consensus for WP and WM) are genuine strengths, and the cross-stimulus correlations with the sneaker but not the dress are informative. The paper is also candid about its limitations, explicitly noting the absence of a no-sock control and the weakness of the experience-to-prior link. The central causal claim about fabric priors, however, is not secured by the reported data, and this affects the abstract's headline assertion.","major_comments":[{"comment":"The abstract and Discussion assert that a pre-existing white-sock fabric prior drives the P2 percept, but the only supporting evidence is the cross-sectional association in Fig. 7, in which the sock-color report was collected immediately after the P2 judgment. With this fixed order, the association is equally compatible with the P2 percept anchoring the subsequent sock report, a shared response bias, or an unmeasured third factor. No counterbalanced order, no-sock control, or manipulation of the illumination-mirroring object is reported. The causal priority of the fabric prior is therefore load-bearing for the paper's central claim but unsupported by the current design.","section":"Fig. 7 and §'What determines the perception of the crocs displays?'"},{"comment":"The Discussion concedes that the no-sock condition was not run (the passage beginning 'it stands to reason that we should create and ask observers about crocs altogether without socks') and that the experience-to-sock link in Fig. 8 is 'obviously less strong than what would be needed to comfortably claim that we understand this aspect.' These concessions directly undercut the abstract's stronger claim that the percept is predicted 'from fabric priors resulting from experience.' The manuscript should either soften the causal claim to a correlational one, explicitly flag the directionality problem in the abstract, or present additional evidence that the prior is causally prior.","section":"Discussion, limitations paragraph"}],"minor_comments":[{"comment":"The caption refers to 'the CTO' in panel C, but the term introduced in the text is 'target colored object' (TCO); this appears to be a typo and should be corrected.","section":"Figure 1 caption"},{"comment":"The statement that 'the exact p-value for either outcome obtained with resampling methods is 0' is not accurate for a bootstrap procedure; report a finite upper bound (e.g., p < 1e-5) instead.","section":"Results, §'Are images of these crocs perceptually ambiguous under typical lighting conditions?'"},{"comment":"The text says 'we used the following seven images of crocs and socks' but Figure 3 also includes the dress and sneaker; please clarify that the seven images refer specifically to the crocs-and-socks displays.","section":"Method, Stimulus design"},{"comment":"The phrase 'amounts to a personality characteristic' is stronger than a single cross-sectional survey can support; 'individual difference measure' would be more appropriate.","section":"Discussion"},{"comment":"The exact order of the dress and sneaker questions relative to the sock-color question is not specified; providing the question sequence would help readers assess potential order effects.","section":"Method, Survey"},{"comment":"A few references (e.g., Land & McCann 1971) lack page numbers or other bibliographic detail; please ensure consistent reference formatting.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The paper's descriptive findings are solid and the design principle is a worthwhile contribution, but the causal claim about fabric priors is currently overstated relative to the evidence. Because the missing controls could be supplied in future work but not within this manuscript, I recommend major revision rather than rejection: the authors should either reframe the central claim as correlational or add the necessary experimental evidence."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things you should know about this paper. First, it's the first time someone has deliberately built categorically ambiguous color displays rather than waiting for one to go viral. That design principle—SURFPAD—is real and the five crocs-and-socks images do produce the kind of large categorical disagreement the dress and sneaker produced by accident. Second, the paper's headline claim that 'we can predict the percept from fabric priors resulting from experience' is not actually established by the data. The survey asks about sock color right after the P2 judgement, so the correlation between seeing the socks as white and seeing the crocs as pink could be reverse causation or a shared response bias. There's no counterbalanced order, no no-sock control, and no manipulation of the IMO, and the authors themselves note in the Discussion that a no-sock condition should be run.\n\nWhat's good: large sample (5,762 after exclusions), sensible bootstrap analysis, and the unambiguous controls (98% agreement on the same crocs under typical light) rule out a lot of low-level or label-based explanations. The relation to the sneaker and dress is also informative, and the age effects are interesting even if hard to interpret. The paper engages honestly with its own limitations—it names the missing no-sock condition explicitly.\n\nThe soft spot is load-bearing for the fabric-prior interpretation, but not for the core existence claim. The SURFPAD principle stands even if the specific mechanism is unconfirmed; the stimuli are ambiguous and the disagreement is real. What isn't proven is that a prior for white socks causes the pink percept. That part is a single-order correlation with a plausible confound, and the self-report link between wearing tube socks and the pink percept is, as the authors admit, weak.\n\nBottom line: this deserves a serious referee. A good referee should push for a cleaner experiment—counterbalanced order, a no-sock or non-white-sock control, and ideally an IMO manipulation—before the causal language is accepted. But as a first demonstration of a general method for generating ambiguous color displays, it's a solid contribution.\n\nI'd cite it for the SURFPAD stimulus set and bring it to reading group; the causality discussion is a good teaching moment.","headline":"Genuinely novel method for creating ambiguous color displays, but the fabric-prior causal claim rests on a single-order correlation with a likely reverse-causation problem.","tokens_in":15465,"tokens_out":2196,"would_cite":true,"duration_ms":22279,"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":"A design principle called SURFPAD lets researchers manufacture color displays that split observers, with a white-sock prior predicting the split.","keywords":["color perception","ambiguous stimuli","SURFPAD","the dress","color constancy","visual priors","perceptual disagreement","fabric priors"],"falsifier":"Run the same survey with the sock-color question asked before any croc display, and include a condition in which the crocs appear with no socks at all. If the white-sock prior causes the pink percept, reversing the order should leave the 12-fold association intact and removing the socks should reduce or eliminate the pink option; if the prior is merely a post-hoc label, the association should shrink or vanish when the order is reversed.","tokens_in":14536,"feed_emoji":"🧦","tokens_out":8157,"duration_ms":72821,"temperature":0.7,"pith_summary":"This paper claims that profoundly ambiguous color displays can be created on demand rather than waiting for rare internet accidents to surface. The authors photograph pink and mint crocs under carefully titrated colored light on a black background, with plain white tube socks included as an anchor that mirrors the illumination. In an online sample of more than 5,700 participants, the same crocs under typical light produced about 98% agreement on the intended color, while five altered displays produced categorical splits—grey, green, pink, or beige—with no single answer near consensus. The paper argues that the split is predicted by whether observers see the socks as white, reflecting a fabric prior built from everyday experience with white tube socks, and it links this mechanism to the dress and the sneaker as different priors driving polarized perception. If right, this turns an internet curiosity into a laboratory tool and a general mechanism for disagreement.","feed_headline":"Five new displays make color perception split on demand","feed_subtitle":"Crocs lit by carefully colored light become ambiguous; a white-sock prior predicts which color people see.","key_machinery":"The load-bearing device is SURFPAD—Substantial Uncertainty combined with Ramified or Forked Priors and Assumptions yields Disagreement—implemented with two named objects: a colored target object (CTO), the croc, whose reflectance is neutralized by titrated light so its pixels sit near grey, and an illumination-mirroring object (IMO), the sock, which reflects the color of the light and is usually white in experience. The trick is that the sock can either be read as white, allowing color-constancy correction back to the manufacturer's color, or as colored, making the image a literal grey croc. That fork in interpretation is what converts one image into opposite percepts.","core_discovery":"The central claim is that a display becomes profoundly color-ambiguous when three conditions are met: the target object is monochromatic and could plausibly be almost any color; the scene removes the background and context cues observers normally use to calibrate color; and an illumination-mirroring object such as a plain sock is included that some observers read as white and others take at face value. Under these conditions, the same pixels are categorically perceived as pink, green, grey, or beige by different observers, and the modal percept can be shifted by choosing the target color and the lighting axis. The paper reports that seeing the socks as white makes the pink percept of the most balanced display more than 12 times more likely, that the same sock prior predicts sneaker perception but not dress perception, and that these differences look like a stable individual characteristic rather than random guessing.","pith_inferences":["If the sock prior is causal, swapping the illumination-mirroring object for a non-white anchor should predictably shift or eliminate the ambiguity; the paper itself names a no-sock control as future work.","The authors' observation that ambiguity tracks the relative position of the croc and sock pixel clouds in hsv space suggests a quantitative recipe: searching for parallel, separated color clouds should generate new ambiguous displays with chosen modal percepts.","The same SURFPAD structure might apply outside color—for example, to auditory or social judgments where one element is assumed neutral by experience—though the paper only gestures in that direction.","Because the sock-color question followed the P2 judgment in the survey, the claimed causal ordering is not yet secured; a reversed-order design is a direct test of that ordering."],"forward_implications":["Because SURFPAD permits creating ambiguous displays on demand, the dress-like phenomenon can be studied in controlled stimulus sets instead of relying on accidental internet images.","The croc displays and the dress and sneaker responses show that disagreement is observer-specific: how someone sees one display predicts how they see the others, implying a stable perceptual disposition.","Sock priors predict croc and sneaker perception but not dress perception, indicating that fabric priors and illumination priors are separate mechanisms operating on different color axes.","Greater confidence in one's beliefs reduces the chance of recovering the color appearance under typical illumination, suggesting that metacognition modulates which prior is applied.","Differential priors under uncertainty, rather than disagreement about the evidence itself, are proposed as the general driver of polarized interpretations beyond color perception."],"supporting_citations":[{"why":"Documents that 'the dress' is a categorically ambiguous color display whose percepts split across observers; supplies the phenomenon SURFPAD aims to reproduce on demand.","marker":"Lafer-Sousa et al., 2015"},{"why":"Shows that illumination priors derived from experience predict dress perception; this is the template for predicting croc percepts from fabric priors.","marker":"Wallisch, 2017a"},{"why":"Argues that implicit assumptions about illumination explain 'the dress', providing the prior-based framework the paper extends.","marker":"Witzel et al., 2017"},{"why":"Retinex theory of lightness and color, used to justify stripping patterned backgrounds and calibration cues from the display.","marker":"Land & McCann, 1971"},{"why":"Demonstrates that patterned chromatic backgrounds shift color appearance, motivating the unpatterned black background in the stimulus design.","marker":"Shevell & Monnier, 2003"},{"why":"Shows that illumination context alters perceived color, motivating the careful titration of complementary light to gray out the crocs.","marker":"Lotto & Purves, 1999"}],"fun_headline_variants":["Sock prior predicts pink vs green in ambiguous crocs display","Crafting color disagreements: crocs and socks reveal priors","White-sock bias drives split color perception in crocs","Designing ambiguity: how sock color primes croc perception","New displays make color perception split via fabric priors"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the sock-color report measures a stable prior that exists before the participant sees the crocs; if asking the sock question after the croc judgment instead creates or shifts that report, the claimed causal direction from white-sock prior to pink-croc percept collapses.","fun_headline_variants_meta":{"raw":{"variants":["Sock prior predicts pink vs green in ambiguous crocs display","Crafting color disagreements: crocs and socks reveal priors","White-sock bias drives split color perception in crocs","Designing ambiguity: how sock color primes croc perception","New displays make color perception split via fabric priors"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000202,"raw_usage":{"total_tokens":1323,"prompt_tokens":830,"completion_tokens":493,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":446,"completion_tokens_details":{"reasoning_tokens":410}},"tokens_in":446,"tokens_out":493,"duration_ms":5905,"temperature":1.0,"reasoning_tokens":410,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:25:43.895744+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same survey with the sock-color question asked before any croc display, and include a condition in which the crocs appear with no socks at all. If the white-sock prior causes the pink percept, reversing the order should leave the 12-fold association intact and removing the socks should reduce or eliminate the pink option; if the prior is merely a post-hoc label, the association should shrink or vanish when the order is reversed.","supporting_citations":[{"cited_title":"the dress","cited_arxiv_id":null,"evidence_quote":"Shows that illumination context alters perceived color, motivating the careful titration of complementary light to gray out the crocs."}],"review_version":1}