{"id":"eb9840a0-630e-4c66-aec2-41563558fed6","arxiv_id":"1908.00661","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"Dynamic luminance and dynamic size produce common-fate grouping like motion, and a 100-participant study ranks motion first, dynamic size second, with context shifting the ordering.","lead":"This paper measures how strongly different visual effects, such as motion, brightness changes, and size changes, make viewers group objects together during animated transitions. It finds motion is generally the strongest grouping cue, but brightness and size changes also group objects, which gives visualization designers new options for guiding attention.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"DL/DS stimuli confound dynamic common fate with static similarity: within-group objects start at identical neutral values and change together, so they are statically similar at every frame; the controlled study never tests the 'not similar at any time t' condition defined in §4.3.2.","rationale":"Good-faith reading: the paper's contribution is an empirical demonstration that common fate generalizes to coordinated luminance/size changes and a quantitative ranking. The decisive condition is that observed grouping choices reflect common dynamic behavior. The current DS/DL stimuli make dynamic groups statically identical at every frame, so the data cannot distinguish common fate from repeated Law of Similarity. This confound is more load-bearing than the reader's parameter-balancing concern, because it attacks H1 itself; if the confound is real, the ranking 'DL > SS', 'DS > SP', and 'DS > DL' are uninterpretable as dynamic grouping strengths. The reader's concern is valid but assumes the dynamic manipulation is already isolated. The proposed control condition directly tests the dissociation. Because the confound can be repaired and partial supporting evidence exists (prior luminance results, qualitative follow-up), I retain the reader's CONDITIONAL verdict rather than escalating to rejection: the paper should not be accepted as-is unless the new experiment or a revised claim addresses this issue. This is why agreement_with_reader is 'disagree'.","tokens_in":20294,"tokens_out":10483,"duration_ms":110002,"concrete_test":"Run a new main-study condition dissociating common change from static similarity for DL and DS. For DL-vs-SS and DS-vs-SP (and ideally all dynamic-vs-static pairs), give within-group objects different starting values but identical signed changes over the transition, e.g., DL group A: 0.40→0.10 and 0.70→0.40, group B: 0.30→0.60 and 0.60→0.90; DS group A: 10→20 and 20→30 px, group B: 30→20 and 20→10 px, keeping all other parameters in Table 2 fixed. No two grouped objects would then share a static luminance or size at any frame. If grouping strength for DL and DS remains above 0.5 in this common-change-only condition, the concern is resolved; if it falls to chance, the main study's dynamic-luminance and dynamic-size effects are attributable to static similarity rather than common fate.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 4.3.2 defines dynamic similarity by |ΔVA−ΔVB|≤θV and states that common fate should hold 'even if objects are not similar at any time t.' The actual stimuli for DL and DS do not instantiate this dissociation. In §5.1 and Table 2, DL groups run 0.5→0.0 and 0.5→1.0, and DS groups run 20→10 and 20→30 pixels, with the two members of each group starting at the identical neutral value and undergoing identical signed changes. Therefore, at every frame during the transition, both members of a group have exactly the same luminance or size, and each group is also statically similar by the Law of Similarity. A participant choosing the darkening or shrinking pair in DL-vs-SS or DS-vs-SP trials could be responding to the static similarity available at each moment, not to common fate. Unlike DP, whose groups are spatially separated and thus not statically similar, DL and DS are never tested in the 'not similar at any time t' condition. This undermines the H1 existence claim, not just the ranking. The §8.2 parameter-balancing limitation is legitimate but secondary; the qualitative follow-up (e.g., P9's 'they didn't start similar... but their rapid growth was similar') offers partial support for common-change grouping, but it does not repair the controlled experiment's confound.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper investigates whether the Gestalt Law of Common Fate extends from coherent motion to coordinated changes in luminance and size. In a crowdsourced experiment, 100 workers completed 12 pairwise competition tasks among three static (position, size, luminance) and three dynamic (motion, luminance change, size change) variables, with three repetitions per task, yielding 3,600 trials. The authors report that dynamic variables dominate static ones, with motion and dynamic size strongest, and that dynamic luminance beats static size. A follow-up qualitative study in a Gapminder-like scatterplot and a U.S. election map suggests that these relative rankings are context-dependent. The paper concludes that dynamic luminance and size create common-fate grouping and discusses implications for animated transitions.","tokens_in":20563,"tokens_out":9494,"duration_ms":93870,"significance":"If the findings withstand scrutiny, the paper would provide a quantitative, crowdsourced extension of the Law of Common Fate beyond velocity, with direct implications for encoding dynamic groups in visualization and for designing animated transitions. The work is thoroughly empirical, uses best-practice estimation with bootstrapped confidence intervals, reports open materials on OSF, and candidly addresses limitations such as parameter selection and context-dependence in the follow-up. The primary weakness is a stimulus confound, detailed below, that bears directly on the H1/H2 claims.","major_comments":[{"comment":"The controlled study does not instantiate the dissociation between dynamic common fate and static similarity that §4.3.2 defines as central to the Law of Common Fate. In all DL trials, both members of a group begin at the identical neutral luminance 0.5 and then undergo the same signed change to 0.0 or 1.0 (Table 2, e.g., SS-DL, SP-DL, DL-DS, DP-DL); in all DS trials, both members begin at the identical radius 20 px and change together to 10 or 30 px (e.g., SS-DS, SP-DS, DL-DS, DP-DS). Consequently, at every frame of the animation the two objects in a group have exactly equal luminance (or size) and are therefore also statically similar under the Law of Similarity. The definition in §4.3.2 explicitly says common fate should hold 'even if objects are not similar at any time t,' but no DL or DS stimulus ever tests this condition: unlike DP, whose grouped objects are spatially separated, no trial contrasts a condition in which objects are statically dissimilar yet dynamically coherent. A participant choosing the darkening or shrinking pair in DL-vs-SS or DS-vs-SP could be responding to static similarity present at every moment, not to common fate. This undermines the H1 existence claim for DL and DS and the H2 claim for DS, not merely the parameter-balancing concern raised in §8.2. I recommend adding a control condition in which group members start from different values and undergo identical signed changes (e.g., luminances 0.3→0.1 vs. 0.7→0.5; radii 14→10 vs. 26→22 px), so that the objects are never statically equal yet have identical delta-V, or, if no new data are collected, substantially narrowing the H1/H2 claims to the conditions actually tested.","section":"§5.1, Table 2, §4.3.2"},{"comment":"The paper's summary ranking 'motion > (dynamic luminance, size, luminance)' is not supported by the data as plotted. Figure 3's left column shows DP and DS both with confidence intervals above 0.5, and the text in Section 8 acknowledges 'no clear distinction between these latter two.' Thus the abstract should not imply a total order with motion strictly above dynamic size. The pairwise contrasts do support a partial order (DP > DL, SS, SL; DS > DL, SP; DL > SS), and the text should present it as such to avoid overstating the evidence.","section":"Section 6.3, Figure 3, abstract"}],"minor_comments":[{"comment":"The ranking notation in the abstract is ambiguous; please use the paper's variable names (DP, DL, DS, SS, SL, SP) consistently and clearly specify which comparisons are being summarized.","section":"Abstract"},{"comment":"The bootstrap analysis aggregates three repetitions into one mean per participant, but the manuscript does not state this explicitly in the methods; please state that the bootstrap resamples participants, not individual trials, so readers do not overestimate precision from the 3,600-trial count.","section":"Section 5.6, Section 6.2"},{"comment":"The footnote about P4 and P5 appears as a stray superscript '5' in the caption; please format it as a proper footnote or note.","section":"Figure 4 caption"},{"comment":"Reference [10] has 'Kbben' (missing an umlaut or diacritic) and reference [32] has 'Visusalization'; these typos should be corrected.","section":"References"},{"comment":"The quotation from Garner that relative grouping strengths are 'fairly impervious to manipulations of discriminability' should be qualified, since the paper itself notes that separability of the dynamic variables is not well characterized; the citation does not directly apply to DL/DS.","section":"Section 8.2"}],"recommendation":"major_revision","confidential_remarks":"The static-similarity confound in the DL/DS stimuli is serious because it touches the main H1/H2 claims; I would not accept the paper in its current form. However, the fix is well-defined (add a condition where group members are never statically equal but share identical changes), and the paper also has substantial strengths: a large crowdsourced design, attention filtering, bootstrapped inference, open materials, and an honest discussion of context-dependence. With that additional control or a suitably narrowed claim, the contribution could be valuable for the visualization community."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague, here's my take on 1908.00661. The paper is worth reading but has a confound that goes beyond the parameter-balancing caveat the authors admit. For the DL and DS conditions, group members start at the exact same neutral value and change identically, so they are statically similar at every frame. The experiment never tests the 'not similar at any time t' case defined in Section 4.3.2. That means the controlled study cannot separate common fate from static similarity for luminance and size. The H1 claim is therefore weaker than presented.\n\nWhat is good: this is the first quantitative comparison of common-fate grouping strength against other Gestalt factors, answering a gap Brooks noted. The large crowdsourced design (3,600 trials, attention filters, bootstrapped CIs) is solid. Extending to dynamic size is genuinely new. The follow-up with Gapminder and election maps is a nice complement, and it honestly shows the ranking shifts, which tempers overgeneralization.\n\nThe parameter concern is real but secondary. The stimulus confound is the load-bearing issue. To fix it, they would need a condition where objects start different but change in parallel, or they should clearly frame the result as about 'congruent change' rather than common fate. The design implication still mostly holds—coordinated luminance or size changes can support grouping—so the paper remains useful for practitioners even after the theoretical claim is cut down.\n\nI would send this to peer review; it is a serious empirical study with a novel comparison. But I would push for major revision to address the confound and reframe the conclusions.","headline":"Useful ranking study for animated transitions, but the DL/DS stimuli never dissociate common fate from static similarity, so the core claim is cleaner in the abstract than in the data.","tokens_in":21109,"tokens_out":3066,"would_cite":true,"duration_ms":30969,"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 paper confirms that the Gestalt Law of Common Fate extends to dynamic luminance and dynamic size, and estimates a partial ranking of their grouping strengths against static position, size, and luminance.","keywords":["common fate","Gestalt psychology","perceptual grouping","animated transitions","visual variables","graphical perception","dynamic luminance","dynamic size"],"falsifier":"Rerun the same four-object conflict task while parametrically varying the magnitudes of dynamic size and luminance changes, equating their discriminability with a staircase procedure; if the dynamic-size-over-luminance ordering reverses or disappears when discriminability is equalized, the ranking is a salience artifact.","tokens_in":20085,"feed_emoji":"👁️","tokens_out":12055,"duration_ms":106610,"temperature":0.7,"pith_summary":"The paper asks whether the Gestalt Law of Common Fate is really about moving together or, more generally, about changing together. A crowdsourced experiment presented four objects divided into two groups in two incompatible ways at once, pitting motion, dynamic luminance, dynamic size, static position, static size, and static luminance against one another. The results confirm that coordinated changes in luminance and size do create perceptual grouping, with motion the strongest cue; the ranking is partial rather than total. A follow-up using an animated scatterplot and a thematic map of election data shows that the relative strengths shift with context, yet people can still read trends from all three dynamic variables. This matters for design because animated transitions can encode dynamic groups using size or luminance changes even when position is already carrying data.","feed_headline":"Objects that grow, shrink, or brighten together are seen as one group","feed_subtitle":"Motion still wins, but changing size or brightness also creates groups; rankings shift in real visualizations.","key_machinery":"The load-bearing instrument is a generalized similarity criterion that turns the Gestalt notion into a measurable quantity. Static similarity holds when two objects' values in a visual variable stay within a threshold, $|V_A(t)-V_B(t)| \\le \\tau_V$; dynamic similarity holds when their changes over a time step stay within a threshold, $|\\Delta V_A(t_{i-1},t_i) - \\Delta V_B(t_{i-1},t_i)| \\le \\theta_V$. Each trial presents four objects with two orthogonal pairwise groupings, each dictated by a different visual variable, and grouping strength is scored by which grouping the observer chooses. This machinery makes the strength of common fate directly comparable, pairwise, against proximity and similarity.","core_discovery":"The paper's central claim is that common fate is not limited to identical velocity: synchronized changes in luminance and in size also group visual objects. In the controlled four-object conflict task, all three dynamic variables had mean grouping strength above the halfway point when competing against static variables, confirming the hypothesis that the law extends to dynamic luminance and dynamic size. The resulting partial ranking is that motion beats dynamic luminance, static size, and static luminance; dynamic size beats dynamic luminance and static position; and dynamic luminance beats static size, with no strict total order across all six variables. In the realistic follow-up, participants used all three dynamic variables to identify trends and to perform conjunction search, but the ranking proved context-dependent: motion dominated in the animated scatterplot, while dynamic luminance and dynamic size were nearly indistinguishable in the election map.","pith_inferences":["The $\\tau_V$ and $\\theta_V$ threshold formalism suggests a direct design rule the paper does not state: choose transition magnitudes so the variable meant to carry a group maximizes its intra-group similarity and inter-group distance relative to competing variables, turning animation design into a constraint-satisfaction problem.","The paper reports that motion can distract attention from size and luminance changes; a testable extension is to cap motion magnitude whenever another dynamic variable is the intended carrier of grouping.","The context-dependence in the follow-up implies that effective grouping strength scales with the number of objects changing in a variable and the magnitude of their change; a controlled experiment that varies cardinality and dynamic range should reproduce the ranking shifts."],"forward_implications":["Animated transitions can encode group membership through simultaneous size or luminance change, not only through shared motion, giving designers extra channels when position is already mapped to data.","To reveal several groups at once, pair variables with similar grouping strengths: dynamic size with static luminance, or dynamic luminance with static position.","To make a single group salient, use the strongest available dynamic variable; for instance, prefer dynamic size over static size when position and luminance are occupied by data.","Viewers can combine up to three dynamic variables in a conjunction search, so multi-variable animated encodings are feasible, but their ranking should be validated in the actual visualization context.","Because the ranking shifted between the controlled and realistic settings, design guidance drawn from controlled perception studies should be re-tested on target visualizations and data."],"supporting_citations":[{"why":"It supplies prior evidence that luminance features varying together over time group together over space, which is the dynamic-luminance effect this paper extends.","marker":"[1]"},{"why":"It demonstrates grouping by common luminance changes, the direct predecessor for the generalized common fate hypothesis.","marker":"[48]"},{"why":"It supplies the original common-fate source, including the passage indicating the principle applies to a wider range of conditions than velocity.","marker":"[64]"},{"why":"It states that no quantitative comparison of common fate grouping strength with other principles existed, which motivates the ranking experiment.","marker":"[12]"},{"why":"It supplies the classical visual-variable framework from which position, size, and luminance were selected as associative, ordered variables.","marker":"[8]"},{"why":"It provides an established accuracy ranking of static graphical encodings, used as a comparison baseline for the dynamic variables.","marker":"[18]"},{"why":"It supports the choice of four objects as within the limit of multiple-object tracking, justifying the task design.","marker":"[14]"},{"why":"It establishes crowdsourced graphical perception as a valid method, grounding the main study's approach.","marker":"[31]"}],"fun_headline_variants":["Common fate goes beyond motion to size and luminance","Synchronized size or brightness changes also group objects","Gestalt grouping by shared size or luminance changes, not just motion","Animated transitions: size and luminance changes trigger grouping","Beyond motion: synchronized size/luminance changes group visuals"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The ranking rests on the assumption that the hand-picked magnitudes of the size, luminance, and motion changes were perceptually balanced across variables; the authors themselves note that the parameter values would have benefited from more principled selection, so if one change was easier to notice than another, the observed ordering could be an artifact of salience rather than intrinsic grouping strength.","fun_headline_variants_meta":{"raw":{"variants":["Common fate goes beyond motion to size and luminance","Synchronized size or brightness changes also group objects","Gestalt grouping by shared size or luminance changes, not just motion","Animated transitions: size and luminance changes trigger grouping","Beyond motion: synchronized size/luminance changes group visuals"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000124,"raw_usage":{"total_tokens":1078,"prompt_tokens":893,"completion_tokens":185,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":509,"completion_tokens_details":{"reasoning_tokens":106}},"tokens_in":509,"tokens_out":185,"duration_ms":2678,"temperature":1.0,"reasoning_tokens":106,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:39:36.798682+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Rerun the same four-object conflict task while parametrically varying the magnitudes of dynamic size and luminance changes, equating their discriminability with a staircase procedure; if the dynamic-size-over-luminance ordering reverses or disappears when discriminability is equalized, the ranking is a salience artifact.","supporting_citations":[{"cited_title":"Alais, R","cited_arxiv_id":null,"evidence_quote":"It supplies prior evidence that luminance features varying together over time group together over space, which is the dynamic-luminance effect this paper extends."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It demonstrates grouping by common luminance changes, the direct predecessor for the generalized common fate hypothesis."},{"cited_title":"Wertheimer","cited_arxiv_id":null,"evidence_quote":"It supplies the original common-fate source, including the passage indicating the principle applies to a wider range of conditions than velocity."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It states that no quantitative comparison of common fate grouping strength with other principles existed, which motivates the ranking experiment."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"It supplies the classical visual-variable framework from which position, size, and luminance were selected as associative, ordered variables."},{"cited_title":"Cavanagh and G","cited_arxiv_id":null,"evidence_quote":"It supports the choice of four objects as within the limit of multiple-object tracking, justifying the task design."}],"review_version":1}