{"id":"a46c72bb-42d5-45b0-b694-3a7a8678715a","arxiv_id":"2510.00824","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":2,"one_line_summary":"VR depth compression makes apertures appear narrower, inflating perceived functional body size; correcting for the vergence-accommodation conflict restores the invariant body-scaled affordance ratio.","lead":"People in virtual reality need wider doorways—both to pass through and to judge that they could pass—than in the real world, and part of this effect is explained by VR's depth compression. After correcting for the vergence-accommodation conflict with a geometric model, the body-scaled relationship between perception and action in VR matches the real world, and the overestimation lingers after leaving VR.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The recovered invariance claim depends on a single unvalidated vergence offset β=0.22° applied uniformly across participants and distances; if the true β for the VIVE Pro 2 differs, the VR–UR equivalence could be a modeling artifact.","rationale":"The reader's weakest assumption correctly identifies the VAC correction model as the load-bearing element. The paper's headline claim—that affordance ratios in VR become equivalent to UR after correcting for VAC—rests entirely on the geometric model of Equations 19–20, with β fixed at 0.22° from the authors' prior work and IPD undisclosed. This is genuinely the most fragile point: the correction is not empirically calibrated within the study, the model assumes no frontoparallel distortion (an assumption the authors themselves flag), and the equivalence tests do not report sensitivity to these parameters. Without a sensitivity analysis or an independent validation of β for the VIVE Pro 2, the recovered invariance could be an artifact of choosing a convenient offset. This concern does not overturn the behavioral findings (elevated thresholds, aftereffect), but it does undermine the mechanistic interpretation and the strong invariance-recovery claim. Since the reader already reached CONDITIONAL with high confidence, my stress-test does not change the verdict; it reinforces the conditionality by specifying a concrete test. I agree with the reader's identification of the same weakest assumption.","tokens_in":22094,"tokens_out":5260,"duration_ms":41576,"concrete_test":"Perform a sensitivity analysis over plausible ranges: vary β from 0° to 0.5° in 0.02° steps and IPD from 5.5 to 7.5 cm, recompute the corrected VR perceptual thresholds and affordance ratios using Equations 19–20, and rerun the equivalence tests (TOST with pre-specified bounds) at each combination. If equivalence with UR holds only for a narrow β/IPD window that excludes the best independent estimate for the VIVE Pro 2, the invariance claim lacks robustness. Additionally, independently measure the depth compression ratio in the same HMD at 2.5 m using a perceptual matching task; if the measured ratio differs from the model's predicted 0.88, the correction is falsified.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 4.2 uses a single fixed vergence offset β=0.22° (from Wang et al., 2024b, under review) to correct all VR perceptual thresholds via Equations 19–20. This correction is the only mechanism that turns the significantly elevated VR affordance ratios (1.35 pre-action, 1.25 post-action) into ratios statistically equivalent to UR (1.12). However, β is never directly measured or validated for the HTC VIVE Pro 2 used in this experiment, and the IPD that enters Equation 18 is not reported. The derivation also explicitly assumes VAC does not affect frontoparallel dimensions (Section 4: 'Assuming the perceptual perturbations imposed by VR do not affect the perceived dimensions on a frontoparallel plane'). If the true vergence offset for this HMD is smaller than 0.22°, or if any lateral compression exists, the corrected VR thresholds would be under-corrected and the equivalence with UR would fail. Conversely, a larger β would over-correct. Because β and IPD are effectively free parameters and no sensitivity analysis is reported, the central claim of 'recovered invariant geometrical scaling' is not yet demonstrated to be a robust perceptual fact rather than an artifact of the chosen correction.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper examines whether immersive VR alters perceived functional body size using a passable-aperture paradigm. In a within-subject design, 60 participants performed action (sidle through) and perception (aperture-width adjustment) tasks in both physical reality (UR) and VR. The authors report significantly higher action and perceptual thresholds in VR, higher affordance ratios (perceptual/action threshold), and a persisting aftereffect in UR after VR exposure. A geometrical model based on the vergence-accommodation conflict (VAC) is then applied to correct VR perceptual thresholds, using a fixed vergence offset β=0.22° taken from prior work. After this correction, the affordance ratios in VR become statistically equivalent to those in UR, which the authors interpret as a recovered invariant geometrical scaling. The paper argues that VR-induced depth compression, not vertical or lateral compression, best explains the disproportionate increase in perceptual thresholds.","tokens_in":22418,"tokens_out":4219,"duration_ms":378501,"significance":"If the central invariance-recovery claim holds, the paper makes a valuable contribution to VR perception and affordance research: it provides a quantitative, theory-driven account of how VAC-induced depth compression affects body-scaled affordance judgments, and it documents a novel aftereffect in unmediated reality. The behavioral effects are large and robust (action threshold F(1,59)=92.94; perceptual threshold F=197.54), the psychometric fitting is appropriate, and the derived depth compression ratio (0.88) is consistent with an independent prior finding (0.84). The modelling is principled and the use of a single external parameter is an empirical strength if that parameter is valid for the present device. However, the central equivalence claim depends on the unvalidated transfer of β to the HTC VIVE Pro 2, the unreported IPD, and an untested frontoparallel assumption, so the paper's significance is conditional on these points being resolved.","major_comments":[{"comment":"The recovered invariance claim rests entirely on the VAC correction with a single fixed vergence offset β = 0.22°, taken from prior work (Wang et al., 2024b; Wang et al., under review). The manuscript does not report the IPD used in Eq. 18, nor does it provide a sensitivity analysis for β. If the true VAC offset for the HTC VIVE Pro 2 differs from 0.22°, the corrected perceptual thresholds—and hence the equivalence with UR in Fig. 6b—would change. Please report the IPD, justify the transfer of β to this specific HMD, and provide a sensitivity analysis sweeping β over a plausible range (e.g., 0.1° to 0.4°) to show that the equivalence conclusion is robust.","section":"Section 4.2, Eqs. 19-20"},{"comment":"The equivalence tests are reported as paired TOST lower/upper t-values with Hedge's g and 90% CIs, but the equivalence margin is never stated. Without a pre-specified and justified margin, 'statistical equivalence' is not interpretable. Moreover, for the UR vs. VR Pre-Action comparison, the 90% CI for Hedge's g is [−0.45, −0.03], which may not fall within a reasonably small equivalence bound (e.g., ±0.2). Please report the equivalence bounds used, justify them, and provide the corresponding confidence interval for the mean difference.","section":"Section 4.2, equivalence tests"},{"comment":"The model assumes that 'the perceptual perturbations imposed by VR do not affect the perceived dimensions on a frontoparallel plane.' This assumption is the linchpin of the depth-compression attribution. The paper argues that lateral compression is unlikely because the virtual environment is visually rich, citing Kelly et al. (2015), but this argument is qualitative. If any frontoparallel compression exists, the corrected aperture widths would be systematically overestimated, and the recovered invariance could be a modelling artifact. Please provide a quantitative test or explicit limitation (e.g., measuring perceived width of a frontoparallel extent in the same setup) or, failing that, temper the central claim.","section":"Section 4, frontoparallel assumption"},{"comment":"The avatar weight manipulation (±20%, n=20 per group) is deliberately ignored in all analyses. Since the avatar is the body whose size is being perceived, pooling across the three manipulation groups could affect the observed affordance ratios and the claimed equivalence. Please include the avatar group as a factor in the repeated-measures ANOVAs (or at least conduct a preliminary analysis demonstrating no group differences) before asserting that the recovered invariance is a general phenomenon.","section":"Sections 3.1-3.2, avatar manipulation"}],"minor_comments":[{"comment":"The text states 'Sixty (60) adults (33 females and 28 males)' but 33+28=61. Please correct the numbers.","section":"Section 2.1, Participants"},{"comment":"The phrase 'only 4 measurements (i.e., height, weight, arm span, fingers, and inseam height)' lists five items. Please clarify which four were used.","section":"Section 2.2, Stimuli and apparatus"},{"comment":"The phrase 'perceived passible aperture threshold' should be 'passable'. Search the text for other instances of 'passible'.","section":"Section 3.1, typo"},{"comment":"Figure 6b would be easier to interpret if the equivalence bounds were drawn or if the raw adjusted thresholds were reported in a supplementary table. Currently the reader cannot verify the equivalence margin from the figure.","section":"Section 4.2, data presentation"},{"comment":"The affordance ratio is defined as perceptual threshold divided by action threshold, but the text in the abstract says 'perceptual threshold over action threshold.' This is consistent; however, please ensure the ratio formula in Eq. 1 and the text always use the same ordering to avoid confusion.","section":"Section 3.2, ratio notation"}],"recommendation":"major_revision","confidential_remarks":"The behavioral results are strong and the modelling approach is interesting, but the central invariance-recovery claim is currently under-specified. The missing IPD, unvalidated transfer of β, absent sensitivity analysis, and unreported equivalence bounds are all fixable within the manuscript's scope. I would be willing to accept conditional on these points being adequately addressed."},"author_rebuttal":null,"desk_editor":null,"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper claims that VR's depth compression—caused by the vergence-accommodation conflict—makes virtual apertures look narrower than depicted, inflating perceived body size; correcting for this distortion recovers the same body-scaled rati","keywords":["virtual reality","affordances","perceived body size","vergence-accommodation conflict","depth compression","passable aperture","body-scaled perception","perception-action calibration"],"falsifier":"Measure each participant's actual vergence offset and interpupillary distance, apply the correction per individual rather than with a fixed 0.22°, and test frontal-plane width perception separately. If per-participant corrections do not make VR affordance ratios equivalent to real-world ratios, or if apertures are also perceived as narrower when viewed front-on (no depth dimension), the central depth-compression explanation fails.","tokens_in":21969,"feed_emoji":"🥽","tokens_out":5783,"duration_ms":52125,"temperature":0.7,"pith_summary":"Virtual reality makes the world look shallower, and the paper shows that this single optical distortion changes how large people feel relative to their surroundings. In a sidle-through-aperture task, participants needed wider gaps in VR to act and judged even wider gaps as necessary; the gap between judgment and action grew beyond what uncertainty alone could explain. The authors trace that extra gap to depth compression from the vergence-accommodation conflict, model it geometrically, and show that after subtracting the distortion the ratio of judged to action width in VR equals the ratio in physical reality. If true, this means the perceptual system stays functionally tuned to action capabilities even inside VR, and the apparent body-size change is a correctable geometrical illusion rather than a failure of body scaling.","feed_headline":"Depth-corrected VR restores real-world body perception","feed_subtitle":"A single vergence offset explains why virtual doorways feel too narrow and why the effect lingers after leaving VR.","key_machinery":"The central object is the affordance ratio, π = perceptual threshold / action threshold, a body-scaled index of whether perceived passability is aligned with actual action capability—the paper's measure of perceived functional body size. The correcting mechanism is a binocular-geometry model of the vergence-accommodation conflict (VAC): because a head-mounted display holds accommodation at one focal distance while vergence varies, an angular offset β is added to the vergence angle, which shifts the perceived location of every point along the line of sight. Solving the geometry with β = 0.22° (Eqs. 19–20) yields the perceived aperture location and width; subtracting this predicted distortion","core_discovery":"Using 60 participants and a passable-aperture paradigm in both physical reality and a head-mounted display, the study finds that action thresholds rise from 28.12 cm to 35.29 cm and perceptual thresholds rise from 31.23 cm to 43–47 cm in VR. The affordance ratio—perceptual threshold divided by action threshold—rises from 1.12 in reality to 1.35 before action calibration and 1.25 after, meaning judgment is inflated beyond what motor uncertainty predicts. The paper attributes the residual inflation to depth compression: the fixed focal distance of the headset biases vergence inward, so an aperture at 2.5 m is perceived near 2.19 m and therefore as narrower. Applying a 0.22° vergence offset in","pith_inferences":["The paper does not test this, but per-participant correction using each person's measured interpupillary distance and vergence offset should remove the residual differences more completely than the fixed 0.22° offset; this could be checked in a replication.","If the distortion is purely optical, rendering the virtual aperture slightly wider (or compressing the scene's scale) by the predicted vergence offset should eliminate the overestimation and possibly the aftereffect; this is a direct design lever the paper does not test.","The same geometric argument should apply to other body-scaled affordances, such as stair height or reachability, where depth compression would shrink apparent distances and shift judgment boundaries; extending the paradigm is a natural next step.","The persistence of the aftereffect raises the possibility that repeated VR sessions build a lasting recalibration of perceived body-environment scaling, which would matter for training and rehabilitation applications; this remains speculative."],"forward_implications":["Users of VR headsets will systematically misjudge passability of openings, needing roughly 5–6 cm extra width in perceptual judgments even after motor uncertainty is accounted for.","The mismatch is not a breakdown of body-scaled perception: once the vergence offset is corrected, perceptual and action thresholds scale the same in VR as in reality.","Experience performing the action in VR partly recalibrates the judgment, but the optical distortion remains until explicitly modeled or removed.","The distortion transfers to the physical world: after VR exposure, participants judged they needed wider real openings, an aftereffect consistent with lingering adaptation of the vergence system.","Viewing distance matters; depth compression grows with distance, explaining why studies that allowed closer viewing found no VR/UR difference, whereas fixed distant viewing shows large differences."],"fun_headline_variants":["Depth fix in VR restores real body perception","VR's depth error makes bodies seem larger","Fixing VR depth distortion brings back accurate body size","Why VR makes you feel wider: depth compression","After VR, body-size overestimation lingers"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The invariance claim rests on a single fixed vergence offset of 0.22° applied to every participant, plus the assumption that VR does not compress frontoparallel (frontal-plane) dimensions; if either assumption fails for the actual headset or individuals, the recovered equality is an artifact of the correction model.","fun_headline_variants_meta":{"raw":{"variants":["Depth fix in VR restores real body perception","VR's depth error makes bodies seem larger","Fixing VR depth distortion brings back accurate body size","Why VR makes you feel wider: depth compression","After VR, body-size overestimation lingers"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000242,"raw_usage":{"total_tokens":1408,"prompt_tokens":834,"completion_tokens":574,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":578,"completion_tokens_details":{"reasoning_tokens":502}},"tokens_in":578,"tokens_out":574,"duration_ms":6166,"temperature":1.0,"reasoning_tokens":502,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-04T13:00:03.511128+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure each participant's actual vergence offset and interpupillary distance, apply the correction per individual rather than with a fixed 0.22°, and test frontal-plane width perception separately. If per-participant corrections do not make VR affordance ratios equivalent to real-world ratios, or if apertures are also perceived as narrower when viewed front-on (no depth dimension), the central depth-compression explanation fails.","supporting_citations":[],"review_version":1}