{
  "paperSlug": "kammer-2026-foveal-feedback",
  "version": 3,
  "synthesizerPrompt": "You are reconstructing the argument of a scientific paper from its decomposed claim structure.\n\nYou have only the claims and the relations between them. You do not have the paper's title, abstract, prose, authors, or interpretive framing. You see the claim sentences, the panels they're tied to, their argumentative role, and the structural relations between them.\n\nThe claim graph carries multiple kinds of relation, each representing a different argumentative move:\n\n- **`requires`** \u2014 A depends on B being true. Mechanistic / hierarchical chain.\n- **`entails` / `derived-from`** \u2014 Hypothesis \u2192 prediction. Deductive entailment.\n- **`tests`** \u2014 Empirical claim \u2192 prediction it tests.\n- **`supports` / `refutes`** \u2014 Empirical claim \u2192 hypothesis it supports or refutes. Abductive inference.\n- **`rules-out`** \u2014 A's evidence eliminates an alternative. Argument by elimination.\n- **`dissociates-with`** \u2014 A and B jointly establish a dissociation. Argument by contrast.\n- **`validates`** \u2014 A is a control or sign-flip that strengthens B. Argument by disconfirmation.\n- **`predicts` / `confirms`** \u2014 predictive validation across model and experiment.\n- **`scopes`** \u2014 A is a boundary condition on B (or on all claims). Argument by qualified scope.\n- **`interprets`** \u2014 A reframes empirical B through theoretical / literature lens. Argument by reframing \u2014 not derivation, but an act of mapping.\n- **`enables-method`** \u2014 A is the methodological capability that warrants B's interpretability.\n\nEach claim has a role: `hypothesis`, `prediction`, `empirical`, `synthesis`, `interpretation`, `methodological`, `control`, or `scope`.\n\nScientific argument typically combines three reasoning forms:\n- **Deduction** \u2014 `entails`/`derived-from` edges.\n- **Induction** \u2014 `requires`/`supports` edges.\n- **Abduction** \u2014 `supports`/`refutes` from empirical back to hypothesis.\n\nYour task: write a paragraph (200\u2013400 words) articulating what this paper is arguing, derived from the structure alone, in the style of a scientific abstract.\n\nUse the right rhetorical move for the right structural relation. Honor epistemic markers and roles. Don't add background framing or literature you don't have. Don't speculate beyond claims. The structure of the argument should be visible in the prose.\n\nOutput:\n1. Synthesis paragraph\n2. Traceback\n\nClaim graph follows.\n\n---\n\n## Hypotheses\n\n- **hypothesis-feedback-carries-shape-not-category** [hypothesis, hypothesis, epistemic=hypothesis, status=unknown]\n  Foveal feedback to early visual cortex carries shape information but not semantic category \u2014 the level of representation appropriate to V1, not LO.\n  - _entails_ **prediction-v1-category-drops-shape-preserved** [prediction, hypothesis, epistemic=prediction, status=unknown]\n    The shape-not-category hypothesis predicts that V1 cross-category decoding should drop under feedback while cross-shape decoding should be preserved.\n    - _tests [supports hypothesis | dissociates-with: lo-shows-reversed-specificity]_ **v1-category-decoding-drops-in-feedback** [empirical, fig3B, epistemic=moderate, status=unverified:compute-infeasible]\n      Within foveal V1, cross-category decoding drops significantly under feedback while cross-shape decoding is preserved, marking shape as the feature dimension carried by the feedback signal.\n    - _tests [supports hypothesis]_ **v2-v3-generalize-shape-not-category** [empirical, fig3-figure-supplement-1, epistemic=moderate, status=unverified:compute-infeasible]\n      The shape-sensitive, category-insensitive feedback profile extends to foveal V2 and V3, locating the effect in early visual cortex broadly rather than in V1 alone.\n  - _entails_ **prediction-lo-inverse-pattern** [prediction, hypothesis, epistemic=prediction, status=unknown]\n    The level-of-representation hypothesis predicts that LO will show the inverse profile to foveal V1 \u2014 cross-shape drops, cross-category preserved.\n    - _tests [supports hypothesis | dissociates-with: v1-category-decoding-drops-in-feedback]_ **lo-shows-reversed-specificity** [empirical, fig3B, epistemic=strong, status=unverified:compute-infeasible]\n      Lateral occipital cortex shows the inverse profile to foveal V1 \u2014 cross-shape decoding drops while cross-category decoding survives \u2014 yielding a double dissociation that rules out a generic sensitivity argument for the V1 result.\n\n- **hypothesis-feedback-not-spillover** [hypothesis, hypothesis, epistemic=hypothesis, status=unknown]\n  The foveal V1 decoding signal reflects genuine top-down feedback rather than passive spillover from peripheral receptive fields.\n  - _entails_ **prediction-u-shape-eccentricity** [prediction, hypothesis, epistemic=prediction, status=unknown]\n    The feedback hypothesis predicts a U-shaped eccentricity profile of decoding accuracy in early visual cortex (parafoveal dip, foveal rise), distinguishable from the monotonic decay predicted by spillover.\n    - _tests [supports hypothesis | rules-out: passive spillover from large peripheral receptive fields]_ **u-shaped-eccentricity-rejects-spillover** [control, fig2B, epistemic=strong, status=unverified:compute-infeasible]\n      Decoding accuracy across V1, V2, and V3 dips at parafoveal eccentricities and rises again at the fovea, a U-shaped profile that is incompatible with passive spillover from large peripheral receptive fields.\n\n- **hypothesis-shared-representational-format** [hypothesis, hypothesis, epistemic=hypothesis, status=unknown]\n  Foveal feedback uses the same representational format as bottom-up sensory drive in foveal V1.\n  - _entails_ **prediction-cross-decoding-generalizes** [prediction, hypothesis, epistemic=prediction, status=unknown]\n    The shared-format hypothesis predicts that classifiers trained on feedback responses should cross-decode to direct-stimulation responses.\n    - _tests [supports hypothesis]_ **cross-decoding-experimental-to-control** [empirical, fig2B, epistemic=strong, status=unverified:compute-infeasible]\n      Classifiers trained on foveal feedback responses generalize to direct foveal stimulation, indicating that feedback uses a representational format shared with bottom-up sensory drive.\n\n## Methodological warrants\n\n- **preregistered-design-validates-mvpa** [methodological, methods, epistemic=moderate, status=verified]\n  The MVPA pipeline, ROI definitions, and statistical tests were preregistered, constraining analytic flexibility for the main decoding results \u2014 though the parametric modulation analysis was excluded from the registered plan.\n  - _enables-method_ **foveal-v1-decodes-peripheral-saccade-target** [empirical, fig2A, epistemic=strong, status=unverified:compute-infeasible]\n    Foveal V1 carries decodable information about peripheral saccade targets even when those targets are extinguished before the eye lands, demonstrating retinotopically anticipatory feedback into early visual cortex.\n  - _enables-method_ **foveal-feedback-below-direct-stimulation** [empirical, fig2A, epistemic=strong, status=unverified:compute-infeasible]\n    The feedback signal in foveal V1 is reliably weaker than the response to direct foveal stimulation, consistent with a low-bandwidth top-down channel rather than a fully reinstated sensory representation.\n  - _enables-method_ **cross-decoding-experimental-to-control** [empirical, fig2B, epistemic=strong, status=unverified:compute-infeasible]\n    Classifiers trained on foveal feedback responses generalize to direct foveal stimulation, indicating that feedback uses a representational format shared with bottom-up sensory drive.\n  - _enables-method_ **decoding-shape-sensitive-not-semantic** [synthesis, fig3, epistemic=moderate, status=unverified:compute-infeasible]\n    Foveal feedback carries low-to-mid-level shape information but not semantic category, identifying the feedback content as visual features of the saccade target rather than its identity.\n  - _enables-method_ **v1-category-decoding-drops-in-feedback** [empirical, fig3B, epistemic=moderate, status=unverified:compute-infeasible]\n    Within foveal V1, cross-category decoding drops significantly under feedback while cross-shape decoding is preserved, marking shape as the feature dimension carried by the feedback signal.\n  - _enables-method_ **lo-shows-reversed-specificity** [empirical, fig3B, epistemic=strong, status=unverified:compute-infeasible]\n    Lateral occipital cortex shows the inverse profile to foveal V1 \u2014 cross-shape decoding drops while cross-category decoding survives \u2014 yielding a double dissociation that rules out a generic sensitivity argument for the V1 result.\n  - _enables-method_ **v2-v3-generalize-shape-not-category** [empirical, fig3-figure-supplement-1, epistemic=moderate, status=unverified:compute-infeasible]\n    The shape-sensitive, category-insensitive feedback profile extends to foveal V2 and V3, locating the effect in early visual cortex broadly rather than in V1 alone.\n  - _enables-method_ **u-shaped-eccentricity-rejects-spillover** [control, fig2B, epistemic=strong, status=unverified:compute-infeasible]\n    Decoding accuracy across V1, V2, and V3 dips at parafoveal eccentricities and rises again at the fovea, a U-shaped profile that is incompatible with passive spillover from large peripheral receptive fields.\n  - _requires_ **preregistration-submitted-after-manuscript** [scope, methods, epistemic=strong, status=verified]\n    The preregistration was uploaded to OSF only after the manuscript was submitted, so its evidentiary weight rests on author-cited website timestamps rather than on a public deposit predating data collection.\n\n- **target-excluded-fovea-in-99pct-saccades** [methodological, fig1C, epistemic=strong, status=unverified:compute-infeasible]\n  The gaze-contingent display extinguished the peripheral target before it crossed into the central 2 degrees of visual angle on 99.27 percent of saccades, ruling out direct foveal stimulation as the source of the decoded signal.\n  - _enables-method_ **foveal-v1-decodes-peripheral-saccade-target** [empirical, fig2A, epistemic=strong, status=unverified:compute-infeasible]\n    Foveal V1 carries decodable information about peripheral saccade targets even when those targets are extinguished before the eye lands, demonstrating retinotopically anticipatory feedback into early visual cortex.\n  - _enables-method_ **decoding-shape-sensitive-not-semantic** [synthesis, fig3, epistemic=moderate, status=unverified:compute-infeasible]\n    Foveal feedback carries low-to-mid-level shape information but not semantic category, identifying the feedback content as visual features of the saccade target rather than its identity.\n  - _enables-method_ **v1-category-decoding-drops-in-feedback** [empirical, fig3B, epistemic=moderate, status=unverified:compute-infeasible]\n    Within foveal V1, cross-category decoding drops significantly under feedback while cross-shape decoding is preserved, marking shape as the feature dimension carried by the feedback signal.\n  - _enables-method_ **v2-v3-generalize-shape-not-category** [empirical, fig3-figure-supplement-1, epistemic=moderate, status=unverified:compute-infeasible]\n    The shape-sensitive, category-insensitive feedback profile extends to foveal V2 and V3, locating the effect in early visual cortex broadly rather than in V1 alone.\n  - _enables-method_ **cross-decoding-experimental-to-control** [empirical, fig2B, epistemic=strong, status=unverified:compute-infeasible]\n    Classifiers trained on foveal feedback responses generalize to direct foveal stimulation, indicating that feedback uses a representational format shared with bottom-up sensory drive.\n  - _enables-method_ **foveal-feedback-below-direct-stimulation** [empirical, fig2A, epistemic=strong, status=unverified:compute-infeasible]\n    The feedback signal in foveal V1 is reliably weaker than the response to direct foveal stimulation, consistent with a low-bandwidth top-down channel rather than a fully reinstated sensory representation.\n  - _enables-method_ **u-shaped-eccentricity-rejects-spillover** [control, fig2B, epistemic=strong, status=unverified:compute-infeasible]\n    Decoding accuracy across V1, V2, and V3 dips at parafoveal eccentricities and rises again at the fovea, a U-shaped profile that is incompatible with passive spillover from large peripheral receptive fields.\n  - _enables-method_ **ips-foveal-effect-reverses-in-control** [control, fig4-figure-supplement-1, epistemic=strong, status=unverified:compute-infeasible]\n    Under direct foveal stimulation the IPS\u2013foveal-decoding correlation reverses sign, confirming that the positive coupling seen during feedback is context-specific and not a generic effect of attention or arousal.\n  - _enables-method_ **ips-candidate-driver-foveal-feedback** [synthesis, fig4B, epistemic=moderate, status=unverified:compute-infeasible]\n    Trial-by-trial intraparietal sulcus activity tracks foveal decoding strength more closely than peripheral decoding strength, nominating IPS as a candidate source of the saccade-locked feedback signal.\n\n## Eliminations / controls\n\n- **fef-lo-nonsignificant-after-correction** [control, fig4B, epistemic=moderate, status=unverified:compute-infeasible]\n  Neither FEF nor LO survives Bonferroni correction in the parametric modulation analysis, leaving IPS as the only ROI specifically coupled to foveal decoding rather than a global brain-state effect.\n  - _rules-out_ FEF as candidate driver of foveal feedback\n  - _rules-out_ LO as candidate driver of foveal feedback\n  - _supports_ **ips-candidate-driver-foveal-feedback** [synthesis, fig4B, epistemic=moderate, status=unverified:compute-infeasible]\n    Trial-by-trial intraparietal sulcus activity tracks foveal decoding strength more closely than peripheral decoding strength, nominating IPS as a candidate source of the saccade-locked feedback signal.\n\n- **ips-foveal-effect-reverses-in-control** [control, fig4-figure-supplement-1, epistemic=strong, status=unverified:compute-infeasible]\n  Under direct foveal stimulation the IPS\u2013foveal-decoding correlation reverses sign, confirming that the positive coupling seen during feedback is context-specific and not a generic effect of attention or arousal.\n  - _rules-out_ generic brain-state or arousal explanation of IPS-foveal-decoding coupling\n  - _validates_ **ips-candidate-driver-foveal-feedback** [synthesis, fig4B, epistemic=moderate, status=unverified:compute-infeasible]\n    Trial-by-trial intraparietal sulcus activity tracks foveal decoding strength more closely than peripheral decoding strength, nominating IPS as a candidate source of the saccade-locked feedback signal.\n  - _supports_ **ips-candidate-driver-foveal-feedback** [synthesis, fig4B, epistemic=moderate, status=unverified:compute-infeasible]\n    Trial-by-trial intraparietal sulcus activity tracks foveal decoding strength more closely than peripheral decoding strength, nominating IPS as a candidate source of the saccade-locked feedback signal.\n\n- **u-shaped-eccentricity-rejects-spillover** [control, fig2B, epistemic=strong, status=unverified:compute-infeasible]\n  Decoding accuracy across V1, V2, and V3 dips at parafoveal eccentricities and rises again at the fovea, a U-shaped profile that is incompatible with passive spillover from large peripheral receptive fields.\n  - _rules-out_ passive spillover from large peripheral receptive fields\n  - _supports_ **foveal-v1-decodes-peripheral-saccade-target** [empirical, fig2A, epistemic=strong, status=unverified:compute-infeasible]\n    Foveal V1 carries decodable information about peripheral saccade targets even when those targets are extinguished before the eye lands, demonstrating retinotopically anticipatory feedback into early visual cortex.\n  - _supports_ **hypothesis-feedback-not-spillover** [hypothesis, hypothesis, epistemic=hypothesis, status=unknown]\n    The foveal V1 decoding signal reflects genuine top-down feedback rather than passive spillover from peripheral receptive fields.\n  - _tests_ **prediction-u-shape-eccentricity** [prediction, hypothesis, epistemic=prediction, status=unknown]\n    The feedback hypothesis predicts a U-shaped eccentricity profile of decoding accuracy in early visual cortex (parafoveal dip, foveal rise), distinguishable from the monotonic decay predicted by spillover.\n\n## Synthesis\n\n- **decoding-shape-sensitive-not-semantic** [synthesis, fig3, epistemic=moderate, status=unverified:compute-infeasible]\n  Foveal feedback carries low-to-mid-level shape information but not semantic category, identifying the feedback content as visual features of the saccade target rather than its identity.\n  - _derived-from_ **v1-category-decoding-drops-in-feedback** [empirical, fig3B, epistemic=moderate, status=unverified:compute-infeasible]\n    Within foveal V1, cross-category decoding drops significantly under feedback while cross-shape decoding is preserved, marking shape as the feature dimension carried by the feedback signal.\n  - _derived-from_ **lo-shows-reversed-specificity** [empirical, fig3B, epistemic=strong, status=unverified:compute-infeasible]\n    Lateral occipital cortex shows the inverse profile to foveal V1 \u2014 cross-shape decoding drops while cross-category decoding survives \u2014 yielding a double dissociation that rules out a generic sensitivity argument for the V1 result.\n  - _derived-from_ **v2-v3-generalize-shape-not-category** [empirical, fig3-figure-supplement-1, epistemic=moderate, status=unverified:compute-infeasible]\n    The shape-sensitive, category-insensitive feedback profile extends to foveal V2 and V3, locating the effect in early visual cortex broadly rather than in V1 alone.\n  - _interprets_ **hypothesis-feedback-carries-shape-not-category** [hypothesis, hypothesis, epistemic=hypothesis, status=unknown]\n    Foveal feedback to early visual cortex carries shape information but not semantic category \u2014 the level of representation appropriate to V1, not LO.\n\n- **ips-candidate-driver-foveal-feedback** [synthesis, fig4B, epistemic=moderate, status=unverified:compute-infeasible]\n  Trial-by-trial intraparietal sulcus activity tracks foveal decoding strength more closely than peripheral decoding strength, nominating IPS as a candidate source of the saccade-locked feedback signal.\n  - _derived-from_ **fef-lo-nonsignificant-after-correction** [control, fig4B, epistemic=moderate, status=unverified:compute-infeasible]\n    Neither FEF nor LO survives Bonferroni correction in the parametric modulation analysis, leaving IPS as the only ROI specifically coupled to foveal decoding rather than a global brain-state effect.\n  - _derived-from_ **ips-foveal-effect-reverses-in-control** [control, fig4-figure-supplement-1, epistemic=strong, status=unverified:compute-infeasible]\n    Under direct foveal stimulation the IPS\u2013foveal-decoding correlation reverses sign, confirming that the positive coupling seen during feedback is context-specific and not a generic effect of attention or arousal.\n  - _requires_ **parametric-modulation-exploratory-not-preregistered** [scope, methods, epistemic=strong, status=verified]\n    The parametric modulation analysis identifying IPS as a feedback driver was explicitly exploratory and outside the preregistered plan, so it carries the weight of a hypothesis-generating result rather than a confirmatory test.\n\n## Dissociations\n\n- **lo-shows-reversed-specificity** \u27f7 **v1-category-decoding-drops-in-feedback** _(dissociates-with)_\n  - **lo-shows-reversed-specificity** [empirical, fig3B, epistemic=strong, status=unverified:compute-infeasible]\n    Lateral occipital cortex shows the inverse profile to foveal V1 \u2014 cross-shape decoding drops while cross-category decoding survives \u2014 yielding a double dissociation that rules out a generic sensitivity argument for the V1 result.\n  - **v1-category-decoding-drops-in-feedback** [empirical, fig3B, epistemic=moderate, status=unverified:compute-infeasible]\n    Within foveal V1, cross-category decoding drops significantly under feedback while cross-shape decoding is preserved, marking shape as the feature dimension carried by the feedback signal.\n\n## Scope qualifiers\n\n- **parametric-modulation-exploratory-not-preregistered** [scope, methods, epistemic=strong, status=verified]\n  The parametric modulation analysis identifying IPS as a feedback driver was explicitly exploratory and outside the preregistered plan, so it carries the weight of a hypothesis-generating result rather than a confirmatory test.\n  - _scopes (qualifies)_ **ips-candidate-driver-foveal-feedback** [synthesis, fig4B, epistemic=moderate, status=unverified:compute-infeasible]\n    Trial-by-trial intraparietal sulcus activity tracks foveal decoding strength more closely than peripheral decoding strength, nominating IPS as a candidate source of the saccade-locked feedback signal.\n  - _scopes (qualifies)_ **fef-lo-nonsignificant-after-correction** [control, fig4B, epistemic=moderate, status=unverified:compute-infeasible]\n    Neither FEF nor LO survives Bonferroni correction in the parametric modulation analysis, leaving IPS as the only ROI specifically coupled to foveal decoding rather than a global brain-state effect.\n  - _scopes (qualifies)_ **ips-foveal-effect-reverses-in-control** [control, fig4-figure-supplement-1, epistemic=strong, status=unverified:compute-infeasible]\n    Under direct foveal stimulation the IPS\u2013foveal-decoding correlation reverses sign, confirming that the positive coupling seen during feedback is context-specific and not a generic effect of attention or arousal.\n\n- **preregistration-submitted-after-manuscript** [scope, methods, epistemic=strong, status=verified]\n  The preregistration was uploaded to OSF only after the manuscript was submitted, so its evidentiary weight rests on author-cited website timestamps rather than on a public deposit predating data collection.\n  - _scopes (qualifies)_ **preregistered-design-validates-mvpa** [methodological, methods, epistemic=moderate, status=verified]\n    The MVPA pipeline, ROI definitions, and statistical tests were preregistered, constraining analytic flexibility for the main decoding results \u2014 though the parametric modulation analysis was excluded from the registered plan.\n\n## Standalone empirical\n\n- **cross-decoding-experimental-to-control** [empirical, fig2B, epistemic=strong, status=unverified:compute-infeasible]\n  Classifiers trained on foveal feedback responses generalize to direct foveal stimulation, indicating that feedback uses a representational format shared with bottom-up sensory drive.\n  - _supports_ **hypothesis-shared-representational-format** [hypothesis, hypothesis, epistemic=hypothesis, status=unknown]\n    Foveal feedback uses the same representational format as bottom-up sensory drive in foveal V1.\n  - _tests_ **prediction-cross-decoding-generalizes** [prediction, hypothesis, epistemic=prediction, status=unknown]\n    The shared-format hypothesis predicts that classifiers trained on feedback responses should cross-decode to direct-stimulation responses.\n\n- **foveal-feedback-below-direct-stimulation** [empirical, fig2A, epistemic=strong, status=unverified:compute-infeasible]\n  The feedback signal in foveal V1 is reliably weaker than the response to direct foveal stimulation, consistent with a low-bandwidth top-down channel rather than a fully reinstated sensory representation.\n\n- **foveal-v1-decodes-peripheral-saccade-target** [empirical, fig2A, epistemic=strong, status=unverified:compute-infeasible]\n  Foveal V1 carries decodable information about peripheral saccade targets even when those targets are extinguished before the eye lands, demonstrating retinotopically anticipatory feedback into early visual cortex.\n\n- **lo-shows-reversed-specificity** [empirical, fig3B, epistemic=strong, status=unverified:compute-infeasible]\n  Lateral occipital cortex shows the inverse profile to foveal V1 \u2014 cross-shape decoding drops while cross-category decoding survives \u2014 yielding a double dissociation that rules out a generic sensitivity argument for the V1 result.\n  - _supports_ **hypothesis-feedback-carries-shape-not-category** [hypothesis, hypothesis, epistemic=hypothesis, status=unknown]\n    Foveal feedback to early visual cortex carries shape information but not semantic category \u2014 the level of representation appropriate to V1, not LO.\n  - _supports_ **decoding-shape-sensitive-not-semantic** [synthesis, fig3, epistemic=moderate, status=unverified:compute-infeasible]\n    Foveal feedback carries low-to-mid-level shape information but not semantic category, identifying the feedback content as visual features of the saccade target rather than its identity.\n  - _tests_ **prediction-lo-inverse-pattern** [prediction, hypothesis, epistemic=prediction, status=unknown]\n    The level-of-representation hypothesis predicts that LO will show the inverse profile to foveal V1 \u2014 cross-shape drops, cross-category preserved.\n  - _dissociates-with_ **v1-category-decoding-drops-in-feedback** [empirical, fig3B, epistemic=moderate, status=unverified:compute-infeasible]\n    Within foveal V1, cross-category decoding drops significantly under feedback while cross-shape decoding is preserved, marking shape as the feature dimension carried by the feedback signal.\n\n- **v1-category-decoding-drops-in-feedback** [empirical, fig3B, epistemic=moderate, status=unverified:compute-infeasible]\n  Within foveal V1, cross-category decoding drops significantly under feedback while cross-shape decoding is preserved, marking shape as the feature dimension carried by the feedback signal.\n  - _supports_ **hypothesis-feedback-carries-shape-not-category** [hypothesis, hypothesis, epistemic=hypothesis, status=unknown]\n    Foveal feedback to early visual cortex carries shape information but not semantic category \u2014 the level of representation appropriate to V1, not LO.\n  - _supports_ **decoding-shape-sensitive-not-semantic** [synthesis, fig3, epistemic=moderate, status=unverified:compute-infeasible]\n    Foveal feedback carries low-to-mid-level shape information but not semantic category, identifying the feedback content as visual features of the saccade target rather than its identity.\n  - _tests_ **prediction-v1-category-drops-shape-preserved** [prediction, hypothesis, epistemic=prediction, status=unknown]\n    The shape-not-category hypothesis predicts that V1 cross-category decoding should drop under feedback while cross-shape decoding should be preserved.\n  - _dissociates-with_ **lo-shows-reversed-specificity** [empirical, fig3B, epistemic=strong, status=unverified:compute-infeasible]\n    Lateral occipital cortex shows the inverse profile to foveal V1 \u2014 cross-shape decoding drops while cross-category decoding survives \u2014 yielding a double dissociation that rules out a generic sensitivity argument for the V1 result.\n\n- **v2-v3-generalize-shape-not-category** [empirical, fig3-figure-supplement-1, epistemic=moderate, status=unverified:compute-infeasible]\n  The shape-sensitive, category-insensitive feedback profile extends to foveal V2 and V3, locating the effect in early visual cortex broadly rather than in V1 alone.\n  - _supports_ **hypothesis-feedback-carries-shape-not-category** [hypothesis, hypothesis, epistemic=hypothesis, status=unknown]\n    Foveal feedback to early visual cortex carries shape information but not semantic category \u2014 the level of representation appropriate to V1, not LO.\n  - _tests_ **prediction-v1-category-drops-shape-preserved** [prediction, hypothesis, epistemic=prediction, status=unknown]\n    The shape-not-category hypothesis predicts that V1 cross-category decoding should drop under feedback while cross-shape decoding should be preserved.\n",
  "synthesis": "Using a gaze-contingent paradigm in which a peripheral saccade target is extinguished before fixation on more than 99% of trials, this work asks whether foveal V1 carries information about a saccade target it never directly receives. The first empirical claim establishes that it does: stimulus identity can be decoded from foveal V1 above chance (fig2A), even though decoding remains well below the level achieved with direct foveal stimulation (fig2A) \u2014 consistent with a low-bandwidth top-down channel rather than full sensory reinstatement. Three hypotheses then structure the remaining argument. First, that the foveal signal reflects genuine top-down feedback rather than passive spillover from peripheral receptive fields; this entails a U-shaped eccentricity profile (parafoveal dip, foveal rise) which is observed across V1, V2, and V3 (fig2B), an outcome that supports the feedback hypothesis and rules out the spillover alternative. Second, that feedback uses a representational format shared with bottom-up sensory drive; this entails that classifiers trained on feedback responses should generalize to direct stimulation, and they do (fig2B). Third, that feedback content matches the representational level of the receiving cortex \u2014 shape, not semantic category. This entails a double prediction: V1 cross-category decoding should drop while cross-shape is preserved (fig3B), and LO should show the inverse profile (fig3B); both obtain, jointly establishing a dissociation between V1 and LO, with V2 and V3 extending the V1 pattern. A synthesis claim consolidates these results into the conclusion that foveal feedback carries low-to-mid-level shape information about the saccade target. An exploratory parametric modulation analysis nominates IPS as a candidate driver: trial-by-trial IPS activity tracks foveal decoding, FEF and LO do not survive correction, and the IPS coupling reverses sign under direct stimulation, ruling out a generic arousal account. The main decoding pipeline is preregistered; the IPS analysis is flagged as exploratory and the preregistration was deposited only after submission, qualifications that the structure carries explicitly.",
  "traceback": [
    {
      "step": "Foveal V1 carries decodable information about an extinguished saccade target",
      "relation": "empirical (fig2A) enabled-by gaze-contingent paradigm (fig1C, methodological) and preregistered MVPA pipeline (methods, methodological)",
      "claims": [
        "foveal-v1-decodes-peripheral-saccade-target",
        "target-excluded-fovea-in-99pct-saccades",
        "preregistered-design-validates-mvpa"
      ]
    },
    {
      "step": "Feedback signal is weaker than direct stimulation (low-bandwidth channel)",
      "relation": "empirical contrast (fig2A)",
      "claims": [
        "foveal-feedback-below-direct-stimulation"
      ]
    },
    {
      "step": "Hypothesis 1 \u2014 feedback, not spillover \u2014 entails U-shape eccentricity prediction; tested and supported, alternative ruled out",
      "relation": "hypothesis entails prediction; control claim tests prediction, supports hypothesis, rules-out spillover",
      "claims": [
        "hypothesis-feedback-not-spillover",
        "prediction-u-shape-eccentricity",
        "u-shaped-eccentricity-rejects-spillover"
      ]
    },
    {
      "step": "Hypothesis 2 \u2014 shared representational format \u2014 entails cross-decoding generalization; supported by empirical cross-decoding result",
      "relation": "hypothesis entails prediction; empirical tests prediction, supports hypothesis",
      "claims": [
        "hypothesis-shared-representational-format",
        "prediction-cross-decoding-generalizes",
        "cross-decoding-experimental-to-control"
      ]
    },
    {
      "step": "Hypothesis 3 \u2014 shape not category, level-matched to receiving cortex \u2014 entails twin predictions for V1 and LO",
      "relation": "hypothesis entails two predictions",
      "claims": [
        "hypothesis-feedback-carries-shape-not-category",
        "prediction-v1-category-drops-shape-preserved",
        "prediction-lo-inverse-pattern"
      ]
    },
    {
      "step": "V1 cross-category drops, cross-shape preserved (and same pattern in V2/V3); LO shows reverse profile \u2014 joint dissociation",
      "relation": "two empirical claims test predictions and support hypothesis; dissociates-with edge between V1 and LO results",
      "claims": [
        "v1-category-decoding-drops-in-feedback",
        "v2-v3-generalize-shape-not-category",
        "lo-shows-reversed-specificity"
      ]
    },
    {
      "step": "Synthesis: foveal feedback carries low-to-mid-level shape, not semantic category",
      "relation": "synthesis derived-from three empirical results; interprets the level-matched-feedback hypothesis",
      "claims": [
        "decoding-shape-sensitive-not-semantic"
      ]
    },
    {
      "step": "IPS nominated as candidate feedback driver; FEF/LO eliminated; sign-reversal in control rules out arousal/brain-state account",
      "relation": "synthesis derived-from two control claims; one rules-out alternative ROIs; other rules-out generic-arousal alternative and validates the synthesis",
      "claims": [
        "ips-candidate-driver-foveal-feedback",
        "fef-lo-nonsignificant-after-correction",
        "ips-foveal-effect-reverses-in-control"
      ]
    },
    {
      "step": "Scope qualifiers: parametric modulation analysis is exploratory and outside preregistration; preregistration deposited only post-submission",
      "relation": "scope claims qualify the IPS synthesis and the preregistration warrant respectively",
      "claims": [
        "parametric-modulation-exploratory-not-preregistered",
        "preregistration-submitted-after-manuscript"
      ]
    }
  ]
}