Feedback of peripheral saccade targets to early foveal cortex
Hypotheses The paper tests whether primary visual cortex receives image-specific feedback about peripheral saccade targets before those targets reach the fovea, whether that feedback carries shape rather than semantic category information, and whether it shares a representational format with direct foveal stimulation.
Claims In a gaze-contingent fMRI paradigm in which peripheral targets disappear before fixation, multivariate classifiers decode target identity from foveal retinotopic V1 despite the absence of foveal stimulation; decoding generalizes across experimental and control conditions, follows a U-shaped eccentricity profile inconsistent with spillover or large receptive fields, and preserves shape but not category information. V1 loses category decoding in the feedback condition while lateral occipital cortex shows the inverse pattern, and trial-wise foveal decodability covaries with activity in the intraparietal sulcus.
Inferences Foveal V1 predicts the features of incoming saccade targets via top-down feedback in a shared representational format with bottom-up stimulation, positioning IPS as a candidate driver and nominating foveal prediction as a mechanism underlying transsaccadic perceptual stability. The shape-not-category profile constrains the feedback to low-to-mid-level visual content rather than abstract object identity.
▸ Summary
▸Hypotheses tested
Foveal feedback to early visual cortex carries shape information but not semantic category — the level of representation appropriate to V1, not LO.
The shape-not-category hypothesis predicts that V1 cross-category decoding should drop under feedback while cross-shape decoding should be preserved.
Tested by
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.
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.
The level-of-representation hypothesis predicts that LO will show the inverse profile to foveal V1 — cross-shape drops, cross-category preserved.
Tested by
Lateral occipital cortex shows the inverse profile to foveal V1 — cross-shape decoding drops while cross-category decoding survives — yielding a double dissociation that rules out a generic sensitivity argument for the V1 result.
The foveal V1 decoding signal reflects genuine top-down feedback rather than passive spillover from peripheral receptive fields.
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.
Tested by
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.
The shared-format hypothesis predicts that classifiers trained on feedback responses should cross-decode to direct-stimulation responses.
Tested by
Classifiers trained on foveal feedback responses generalize to direct foveal stimulation, indicating that feedback uses a representational format shared with bottom-up sensory drive.
▸Dissociations
Lateral occipital cortex shows the inverse profile to foveal V1 — cross-shape decoding drops while cross-category decoding survives — yielding a double dissociation that rules out a generic sensitivity argument for the V1 result.
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.
▸Eliminations & validating controls
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.
Under direct foveal stimulation the IPS–foveal-decoding correlation reverses sign, confirming that the positive coupling seen during feedback is context-specific and not a generic effect of attention or arousal.
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.
▸Synthesis claims
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.
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.
▸Standalone empirical findings
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.
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.
▸Methodological warrants
The MVPA pipeline, ROI definitions, and statistical tests were preregistered, constraining analytic flexibility for the main decoding results — though the parametric modulation analysis was excluded from the registered plan.
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.
▸Scope qualifiers
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.
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.
▸All claims (alphabetical)
- cross-decoding-experimental-to-control fig2B
- decoding-shape-sensitive-not-semantic fig3
- fef-lo-nonsignificant-after-correction fig4B
- foveal-feedback-below-direct-stimulation fig2A
- foveal-v1-decodes-peripheral-saccade-target fig2A
- hypothesis-feedback-carries-shape-not-category hypothesis
- hypothesis-feedback-not-spillover hypothesis
- hypothesis-shared-representational-format hypothesis
- interprets-williams-2008-peripheral-spillover hypothesis (spillover alternative), fig2B (U-shape test)
- ips-candidate-driver-foveal-feedback fig4B
- ips-foveal-effect-reverses-in-control fig4-figure-supplement-1
- lo-shows-reversed-specificity fig3B
- parametric-modulation-exploratory-not-preregistered methods
- prediction-cross-decoding-generalizes hypothesis
- prediction-lo-inverse-pattern hypothesis
- prediction-u-shape-eccentricity hypothesis
- prediction-v1-category-drops-shape-preserved hypothesis
- preregistered-design-validates-mvpa methods
- preregistration-submitted-after-manuscript methods
- target-excluded-fovea-in-99pct-saccades fig1C
- u-shaped-eccentricity-rejects-spillover fig2B
- v1-category-decoding-drops-in-feedback fig3B
- v2-v3-generalize-shape-not-category fig3-figure-supplement-1
Abstract mapped to claims
The paper's abstract is shown with each sentence linked to the claim(s) it represents in the dependency graph. Hover or click a sentence to highlight the corresponding claim cards. Below: what the graph contains that the abstract leaves out, and vice versa.
1Human vision is characterized by frequent eye movements and constant shifts in visual input, yet our perception of the world remains remarkably stable. 2Here, we directly demonstrate image-specific foveal feedback to primary visual cortex in the context of saccadic eye movements. 3To this end, we used a gaze-contingent fMRI paradigm, in which peripheral saccade targets disappeared before they could be fixated. 4Despite no direct foveal stimulation, we were able to decode peripheral saccade targets from foveal retinotopic areas, demonstrating that image-specific feedback during saccade preparation may underlie this effect. 5Decoding was sensitive to shape but not semantic category of natural images, indicating feedback of only low-to-mid-level information. 6Cross-decoding to a control condition with foveal stimulus presentation indicates a shared representational format between foveal feedback and direct stimulation. 7Moreover, eccentricity-dependent analyses showed a U-shaped decoding curve, confirming that these results are not explained by spillover of peripheral activity or large receptive fields. 8Finally, fluctuations in foveal decodability covaried with activity in the intraparietal sulcus, thus providing a candidate region for driving foveal feedback. 9These findings suggest that foveal cortex predicts the features of incoming stimuli through feedback from higher cortical areas, which offers a candidate mechanism underlying stable perception.
- M1 preregistered-design-validates-mvpa methods The MVPA pipeline, ROI definitions, and statistical tests were preregistered, constraining analytic flexibility for the main decoding results — though the parametric modulation analysis was excluded from the registered plan.
- Sc2 preregistration-submitted-after-manuscript methods 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.
- E1 foveal-feedback-below-direct-stimulation fig2A 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.
- H1.P1.1 lo-shows-reversed-specificity fig3B Lateral occipital cortex shows the inverse profile to foveal V1 — cross-shape decoding drops while cross-category decoding survives — yielding a double dissociation that rules out a generic sensitivity argument for the V1 result.
- H1.P2.2 v2-v3-generalize-shape-not-category fig3-figure-supplement-1 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.
- Sc1 parametric-modulation-exploratory-not-preregistered methods 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.
- C1 fef-lo-nonsignificant-after-correction fig4B 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.
- C2 ips-foveal-effect-reverses-in-control fig4-figure-supplement-1 Under direct foveal stimulation the IPS–foveal-decoding correlation reverses sign, confirming that the positive coupling seen during feedback is context-specific and not a generic effect of attention or arousal.
- [9] These findings suggest that foveal cortex predicts the features of incoming stimuli through feedback from higher cortical areas, which offers a candidate mechanism underlying stable perception.
Argument from the graphv3
An LLM was given only this paper's enriched claim graph — claims, panel references, roles, and the relations between them — with no access to the abstract, paper prose, or any external context. It was asked to reconstruct the argument. Below: the reconstructed argument, then a comparison to the published abstract above. The two prompts that produced these texts are at the bottom, available for inspection.
Reconstructed argument
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) — 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 — 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.
Compared to the published abstract
Both articulations land on the same five empirical pillars — foveal V1 decoding of the extinguished saccade target, the U-shape eccentricity profile against spillover, cross-decoding to direct stimulation, the shape-not-category specificity, and the IPS coupling — but they make those pillars do very different argumentative work. The v3 synthesis, drawing on the explicit derived-from edges, organizes the prose as three hypothesis-test loops with named alternatives ruled out, an explicit V1/LO double dissociation, and a candidate-elimination story for IPS in which FEF and LO are tested and fail. It marks the synthesis claim that consolidates shape-not-category as a separate interpretive act, and it carries both scope qualifiers — the parametric modulation as exploratory, the preregistration deposited only post-submission. The abstract collapses each hypothesis-test loop to a single result assertion, frames IPS as a positive finding by covariation, omits the LO double dissociation and V2/V3 generalization, suppresses the low-bandwidth amplitude caveat, and carries no scope qualifications. It also opens with perceptual-stability framing and closes with a predictive-coding interpretation, neither of which any graph claim asserts. The graph preserves the argument; the abstract preserves the conclusions and recruits framing the graph cannot supply.
The IPS sentence is again the most diagnostic divergence, and the v3 synthesis makes the underlying inferential shape unmistakable. The abstract reports that 'fluctuations in foveal decodability covaried with activity in the intraparietal sulcus, thus providing a candidate region for driving foveal feedback' — a clean positive-result framing. The v3 graph encodes ips-candidate-driver-foveal-feedback as a synthesis node with two derived-from edges (fef-lo-nonsignificant-after-correction, which itself carries rules-out edges naming FEF and LO as failed candidates; and ips-foveal-effect-reverses-in-control, which carries both a rules-out edge against generic arousal and a validates edge), plus a requires edge to the parametric-modulation-exploratory scope claim. The synthesis prose respects all four edges in a single sentence: IPS as nominated, FEF/LO eliminated, sign-reversal as validation, exploratory status as scope. The abstract preserves the surface claim while stripping the elimination structure, the validation control, and the scope qualifier — three argumentative moves that together transform IPS from a positive discovery into a survivor of elimination, validated by a disconfirming control, bounded by exploratory status. The derived-from edges are doing exactly the work they were designed to do: they make the inferential ancestry of a synthesis claim visible in the prose.
The v3 synthesis cleanly recovers the hypothetico-deductive structure encoded by the enriched edges. Each of the three hypothesis-test loops is rendered explicitly: the hypothesis is named, the prediction is derived with conditional language ('this entails X'), the empirical observation that tests it is reported, and where applicable the alternative ruled out is named. The shape-not-category loop additionally surfaces both predictions, the double dissociation (dissociates-with), and the V2/V3 scope extension. The IPS section reads as an elimination argument followed by a validation control, exactly matching the derived-from / rules-out / validates edge pattern. The synthesis claim that consolidates shape-not-category is treated as a separate inferential step rather than merged into the empirical observation, which honors the derived-from edges from three empirical findings to a synthesis node and the interprets edge to the underlying hypothesis. Scope qualifications are surfaced at the end rather than suppressed. The abstract, by contrast, flattens every hypothesis-test loop into a bare result assertion, reframes the IPS elimination as a positive finding, and omits all scope. The new edges are doing real work: v3 reads as scientific reasoning, where the abstract reads as a list of conclusions.
▸ Show traceback (9 synthesis sentences)
▸ Show structured comparison table (12 rows)
| Kind | Abstract sentence | Synthesis sentence | Category | Note |
|---|---|---|---|---|
| abstract only | Human vision is characterized by frequent eye movements and constant shifts in visual input, yet our perception of the world remains remarkably stable. | background | Generic perceptual-stability framing. No claim in the graph asserts perceptual stability as a phenomenon — pure puzzle-setup. | |
| aligned | Here, we directly demonstrate image-specific foveal feedback to primary visual cortex in the context of saccadic eye movements. | 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. | Both lead with the headline thesis. The abstract asserts ('demonstrate'); the v3 synthesis frames it as a question and folds the methodological enables-method warrant directly into the lead. | |
| aligned | To this end, we used a gaze-contingent fMRI paradigm, in which peripheral saccade targets disappeared before they could be fixated. | 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. | Same methodological fact. The synthesis carries the quantitative validation (99% / 99.27%) the abstract leaves implicit. | |
| reframed | Despite no direct foveal stimulation, we were able to decode peripheral saccade targets from foveal retinotopic areas, demonstrating that image-specific feedback during saccade preparation may underlie this effect. | The first empirical claim establishes that it does: stimulus identity can be decoded from foveal V1 above chance, even though decoding remains well below the level achieved with direct foveal stimulation — consistent with a low-bandwidth top-down channel rather than full sensory reinstatement. | Same headline result, but the v3 synthesis pairs it with the amplitude caveat (foveal-feedback-below-direct-stimulation) — a low-bandwidth qualifier the abstract suppresses. The abstract adds the saccade-preparation gloss, which no claim asserts. | |
| synthesis only | Three hypotheses then structure the remaining argument. | Bridge sentence exposing the hypothetico-deductive structure encoded by the entails edges. The abstract has no analogue — the deductive scaffolding is invisible. | ||
| reframed | Moreover, eccentricity-dependent analyses showed a U-shaped decoding curve, confirming that these results are not explained by spillover of peripheral activity or large receptive fields. | 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, an outcome that supports the feedback hypothesis and rules out the spillover alternative. | Same control argument. Synthesis articulates it as hypothesis → prediction → confirmation → ruled-out alternative; abstract collapses to one assertion. The synthesis names V1/V2/V3 explicitly. | |
| reframed | Cross-decoding to a control condition with foveal stimulus presentation indicates a shared representational format between foveal feedback and direct stimulation. | 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. | Same finding. The synthesis exposes the if-then-observation deductive structure; the abstract states only the conclusion. | |
| reframed | Decoding was sensitive to shape but not semantic category of natural images, indicating feedback of only low-to-mid-level information. | Third, that feedback content matches the representational level of the receiving cortex — shape, not semantic category. This entails a double prediction: V1 cross-category decoding should drop while cross-shape is preserved, and LO should show the inverse profile; both obtain, jointly establishing a dissociation between V1 and LO, with V2 and V3 extending the V1 pattern. | Most striking inferential reframe. The abstract collapses the dissociation to one declarative line. The v3 synthesis articulates the hypothesis, both predictions, the V1/LO double dissociation (dissociates-with), and the V2/V3 scope extension — four distinct evidentiary moves the abstract folds into one. | |
| synthesis only | A synthesis claim consolidates these results into the conclusion that foveal feedback carries low-to-mid-level shape information about the saccade target. | The synthesis-role claim node, marked as derived-from three empirical results and interprets the shape-not-category hypothesis. The abstract folds this conclusion into the dissociation sentence; the synthesis preserves the consolidation as a separate inferential step. | ||
| reframed | Finally, fluctuations in foveal decodability covaried with activity in the intraparietal sulcus, thus providing a candidate region for driving foveal feedback. | 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. | Same finding, opposite epistemic posture. The abstract: positive discovery via covariation. The synthesis: survivor of elimination (FEF, LO ruled out) plus disconfirmation control (sign reversal validates). The elimination and validation structures are invisible in the abstract; 'exploratory' scope is explicit in synthesis. | |
| synthesis only | 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. | Both scope claims (parametric-modulation-exploratory-not-preregistered, preregistration-submitted-after-manuscript) surfaced via scopes/requires edges. The abstract carries neither caveat. | ||
| abstract only | These findings suggest that foveal cortex predicts the features of incoming stimuli through feedback from higher cortical areas, which offers a candidate mechanism underlying stable perception. | interpretive-leap | Predictive-coding framing and the link back to perceptual stability. No claim in the graph asserts this mechanism or this link to stable perception. Pure interpretive bridge to literature the graph does not carry. |
▸ Show synthesizer prompt
You are reconstructing the argument of a scientific paper from its decomposed claim structure.
You 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.
The claim graph carries multiple kinds of relation, each representing a different argumentative move:
- **`requires`** — A depends on B being true. Mechanistic / hierarchical chain.
- **`entails` / `derived-from`** — Hypothesis → prediction. Deductive entailment.
- **`tests`** — Empirical claim → prediction it tests.
- **`supports` / `refutes`** — Empirical claim → hypothesis it supports or refutes. Abductive inference.
- **`rules-out`** — A's evidence eliminates an alternative. Argument by elimination.
- **`dissociates-with`** — A and B jointly establish a dissociation. Argument by contrast.
- **`validates`** — A is a control or sign-flip that strengthens B. Argument by disconfirmation.
- **`predicts` / `confirms`** — predictive validation across model and experiment.
- **`scopes`** — A is a boundary condition on B (or on all claims). Argument by qualified scope.
- **`interprets`** — A reframes empirical B through theoretical / literature lens. Argument by reframing — not derivation, but an act of mapping.
- **`enables-method`** — A is the methodological capability that warrants B's interpretability.
Each claim has a role: `hypothesis`, `prediction`, `empirical`, `synthesis`, `interpretation`, `methodological`, `control`, or `scope`.
Scientific argument typically combines three reasoning forms:
- **Deduction** — `entails`/`derived-from` edges.
- **Induction** — `requires`/`supports` edges.
- **Abduction** — `supports`/`refutes` from empirical back to hypothesis.
Your task: write a paragraph (200–400 words) articulating what this paper is arguing, derived from the structure alone, in the style of a scientific abstract.
Use 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.
Output:
1. Synthesis paragraph
2. Traceback
Claim graph follows.
---
## Hypotheses
- **hypothesis-feedback-carries-shape-not-category** [hypothesis, hypothesis, epistemic=hypothesis, status=unknown]
Foveal feedback to early visual cortex carries shape information but not semantic category — the level of representation appropriate to V1, not LO.
- _entails_ **prediction-v1-category-drops-shape-preserved** [prediction, hypothesis, epistemic=prediction, status=unknown]
The shape-not-category hypothesis predicts that V1 cross-category decoding should drop under feedback while cross-shape decoding should be preserved.
- _tests [supports hypothesis | dissociates-with: lo-shows-reversed-specificity]_ **v1-category-decoding-drops-in-feedback** [empirical, fig3B, epistemic=moderate, status=unverified:compute-infeasible]
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.
- _tests [supports hypothesis]_ **v2-v3-generalize-shape-not-category** [empirical, fig3-figure-supplement-1, epistemic=moderate, status=unverified:compute-infeasible]
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.
- _entails_ **prediction-lo-inverse-pattern** [prediction, hypothesis, epistemic=prediction, status=unknown]
The level-of-representation hypothesis predicts that LO will show the inverse profile to foveal V1 — cross-shape drops, cross-category preserved.
- _tests [supports hypothesis | dissociates-with: v1-category-decoding-drops-in-feedback]_ **lo-shows-reversed-specificity** [empirical, fig3B, epistemic=strong, status=unverified:compute-infeasible]
Lateral occipital cortex shows the inverse profile to foveal V1 — cross-shape decoding drops while cross-category decoding survives — yielding a double dissociation that rules out a generic sensitivity argument for the V1 result.
- **hypothesis-feedback-not-spillover** [hypothesis, hypothesis, epistemic=hypothesis, status=unknown]
The foveal V1 decoding signal reflects genuine top-down feedback rather than passive spillover from peripheral receptive fields.
- _entails_ **prediction-u-shape-eccentricity** [prediction, hypothesis, epistemic=prediction, status=unknown]
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.
- _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]
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.
- **hypothesis-shared-representational-format** [hypothesis, hypothesis, epistemic=hypothesis, status=unknown]
Foveal feedback uses the same representational format as bottom-up sensory drive in foveal V1.
- _entails_ **prediction-cross-decoding-generalizes** [prediction, hypothesis, epistemic=prediction, status=unknown]
The shared-format hypothesis predicts that classifiers trained on feedback responses should cross-decode to direct-stimulation responses.
- _tests [supports hypothesis]_ **cross-decoding-experimental-to-control** [empirical, fig2B, epistemic=strong, status=unverified:compute-infeasible]
Classifiers trained on foveal feedback responses generalize to direct foveal stimulation, indicating that feedback uses a representational format shared with bottom-up sensory drive.
## Methodological warrants
- **preregistered-design-validates-mvpa** [methodological, methods, epistemic=moderate, status=verified]
The MVPA pipeline, ROI definitions, and statistical tests were preregistered, constraining analytic flexibility for the main decoding results — though the parametric modulation analysis was excluded from the registered plan.
- _enables-method_ **foveal-v1-decodes-peripheral-saccade-target** [empirical, fig2A, epistemic=strong, status=unverified:compute-infeasible]
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.
- _enables-method_ **foveal-feedback-below-direct-stimulation** [empirical, fig2A, epistemic=strong, status=unverified:compute-infeasible]
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.
- _enables-method_ **cross-decoding-experimental-to-control** [empirical, fig2B, epistemic=strong, status=unverified:compute-infeasible]
Classifiers trained on foveal feedback responses generalize to direct foveal stimulation, indicating that feedback uses a representational format shared with bottom-up sensory drive.
- _enables-method_ **decoding-shape-sensitive-not-semantic** [synthesis, fig3, epistemic=moderate, status=unverified:compute-infeasible]
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.
- _enables-method_ **v1-category-decoding-drops-in-feedback** [empirical, fig3B, epistemic=moderate, status=unverified:compute-infeasible]
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.
- _enables-method_ **lo-shows-reversed-specificity** [empirical, fig3B, epistemic=strong, status=unverified:compute-infeasible]
Lateral occipital cortex shows the inverse profile to foveal V1 — cross-shape decoding drops while cross-category decoding survives — yielding a double dissociation that rules out a generic sensitivity argument for the V1 result.
- _enables-method_ **v2-v3-generalize-shape-not-category** [empirical, fig3-figure-supplement-1, epistemic=moderate, status=unverified:compute-infeasible]
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.
- _enables-method_ **u-shaped-eccentricity-rejects-spillover** [control, fig2B, epistemic=strong, status=unverified:compute-infeasible]
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.
- _requires_ **preregistration-submitted-after-manuscript** [scope, methods, epistemic=strong, status=verified]
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.
- **target-excluded-fovea-in-99pct-saccades** [methodological, fig1C, epistemic=strong, status=unverified:compute-infeasible]
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.
- _enables-method_ **foveal-v1-decodes-peripheral-saccade-target** [empirical, fig2A, epistemic=strong, status=unverified:compute-infeasible]
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.
- _enables-method_ **decoding-shape-sensitive-not-semantic** [synthesis, fig3, epistemic=moderate, status=unverified:compute-infeasible]
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.
- _enables-method_ **v1-category-decoding-drops-in-feedback** [empirical, fig3B, epistemic=moderate, status=unverified:compute-infeasible]
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.
- _enables-method_ **v2-v3-generalize-shape-not-category** [empirical, fig3-figure-supplement-1, epistemic=moderate, status=unverified:compute-infeasible]
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.
- _enables-method_ **cross-decoding-experimental-to-control** [empirical, fig2B, epistemic=strong, status=unverified:compute-infeasible]
Classifiers trained on foveal feedback responses generalize to direct foveal stimulation, indicating that feedback uses a representational format shared with bottom-up sensory drive.
- _enables-method_ **foveal-feedback-below-direct-stimulation** [empirical, fig2A, epistemic=strong, status=unverified:compute-infeasible]
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.
- _enables-method_ **u-shaped-eccentricity-rejects-spillover** [control, fig2B, epistemic=strong, status=unverified:compute-infeasible]
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.
- _enables-method_ **ips-foveal-effect-reverses-in-control** [control, fig4-figure-supplement-1, epistemic=strong, status=unverified:compute-infeasible]
Under direct foveal stimulation the IPS–foveal-decoding correlation reverses sign, confirming that the positive coupling seen during feedback is context-specific and not a generic effect of attention or arousal.
- _enables-method_ **ips-candidate-driver-foveal-feedback** [synthesis, fig4B, epistemic=moderate, status=unverified:compute-infeasible]
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.
## Eliminations / controls
- **fef-lo-nonsignificant-after-correction** [control, fig4B, epistemic=moderate, status=unverified:compute-infeasible]
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.
- _rules-out_ FEF as candidate driver of foveal feedback
- _rules-out_ LO as candidate driver of foveal feedback
- _supports_ **ips-candidate-driver-foveal-feedback** [synthesis, fig4B, epistemic=moderate, status=unverified:compute-infeasible]
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.
- **ips-foveal-effect-reverses-in-control** [control, fig4-figure-supplement-1, epistemic=strong, status=unverified:compute-infeasible]
Under direct foveal stimulation the IPS–foveal-decoding correlation reverses sign, confirming that the positive coupling seen during feedback is context-specific and not a generic effect of attention or arousal.
- _rules-out_ generic brain-state or arousal explanation of IPS-foveal-decoding coupling
- _validates_ **ips-candidate-driver-foveal-feedback** [synthesis, fig4B, epistemic=moderate, status=unverified:compute-infeasible]
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.
- _supports_ **ips-candidate-driver-foveal-feedback** [synthesis, fig4B, epistemic=moderate, status=unverified:compute-infeasible]
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.
- **u-shaped-eccentricity-rejects-spillover** [control, fig2B, epistemic=strong, status=unverified:compute-infeasible]
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.
- _rules-out_ passive spillover from large peripheral receptive fields
- _supports_ **foveal-v1-decodes-peripheral-saccade-target** [empirical, fig2A, epistemic=strong, status=unverified:compute-infeasible]
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.
- _supports_ **hypothesis-feedback-not-spillover** [hypothesis, hypothesis, epistemic=hypothesis, status=unknown]
The foveal V1 decoding signal reflects genuine top-down feedback rather than passive spillover from peripheral receptive fields.
- _tests_ **prediction-u-shape-eccentricity** [prediction, hypothesis, epistemic=prediction, status=unknown]
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.
## Synthesis
- **decoding-shape-sensitive-not-semantic** [synthesis, fig3, epistemic=moderate, status=unverified:compute-infeasible]
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.
- _derived-from_ **v1-category-decoding-drops-in-feedback** [empirical, fig3B, epistemic=moderate, status=unverified:compute-infeasible]
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.
- _derived-from_ **lo-shows-reversed-specificity** [empirical, fig3B, epistemic=strong, status=unverified:compute-infeasible]
Lateral occipital cortex shows the inverse profile to foveal V1 — cross-shape decoding drops while cross-category decoding survives — yielding a double dissociation that rules out a generic sensitivity argument for the V1 result.
- _derived-from_ **v2-v3-generalize-shape-not-category** [empirical, fig3-figure-supplement-1, epistemic=moderate, status=unverified:compute-infeasible]
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.
- _interprets_ **hypothesis-feedback-carries-shape-not-category** [hypothesis, hypothesis, epistemic=hypothesis, status=unknown]
Foveal feedback to early visual cortex carries shape information but not semantic category — the level of representation appropriate to V1, not LO.
- **ips-candidate-driver-foveal-feedback** [synthesis, fig4B, epistemic=moderate, status=unverified:compute-infeasible]
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.
- _derived-from_ **fef-lo-nonsignificant-after-correction** [control, fig4B, epistemic=moderate, status=unverified:compute-infeasible]
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.
- _derived-from_ **ips-foveal-effect-reverses-in-control** [control, fig4-figure-supplement-1, epistemic=strong, status=unverified:compute-infeasible]
Under direct foveal stimulation the IPS–foveal-decoding correlation reverses sign, confirming that the positive coupling seen during feedback is context-specific and not a generic effect of attention or arousal.
- _requires_ **parametric-modulation-exploratory-not-preregistered** [scope, methods, epistemic=strong, status=verified]
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.
## Dissociations
- **lo-shows-reversed-specificity** ⟷ **v1-category-decoding-drops-in-feedback** _(dissociates-with)_
- **lo-shows-reversed-specificity** [empirical, fig3B, epistemic=strong, status=unverified:compute-infeasible]
Lateral occipital cortex shows the inverse profile to foveal V1 — cross-shape decoding drops while cross-category decoding survives — yielding a double dissociation that rules out a generic sensitivity argument for the V1 result.
- **v1-category-decoding-drops-in-feedback** [empirical, fig3B, epistemic=moderate, status=unverified:compute-infeasible]
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.
## Scope qualifiers
- **parametric-modulation-exploratory-not-preregistered** [scope, methods, epistemic=strong, status=verified]
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.
- _scopes (qualifies)_ **ips-candidate-driver-foveal-feedback** [synthesis, fig4B, epistemic=moderate, status=unverified:compute-infeasible]
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.
- _scopes (qualifies)_ **fef-lo-nonsignificant-after-correction** [control, fig4B, epistemic=moderate, status=unverified:compute-infeasible]
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.
- _scopes (qualifies)_ **ips-foveal-effect-reverses-in-control** [control, fig4-figure-supplement-1, epistemic=strong, status=unverified:compute-infeasible]
Under direct foveal stimulation the IPS–foveal-decoding correlation reverses sign, confirming that the positive coupling seen during feedback is context-specific and not a generic effect of attention or arousal.
- **preregistration-submitted-after-manuscript** [scope, methods, epistemic=strong, status=verified]
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.
- _scopes (qualifies)_ **preregistered-design-validates-mvpa** [methodological, methods, epistemic=moderate, status=verified]
The MVPA pipeline, ROI definitions, and statistical tests were preregistered, constraining analytic flexibility for the main decoding results — though the parametric modulation analysis was excluded from the registered plan.
## Standalone empirical
- **cross-decoding-experimental-to-control** [empirical, fig2B, epistemic=strong, status=unverified:compute-infeasible]
Classifiers trained on foveal feedback responses generalize to direct foveal stimulation, indicating that feedback uses a representational format shared with bottom-up sensory drive.
- _supports_ **hypothesis-shared-representational-format** [hypothesis, hypothesis, epistemic=hypothesis, status=unknown]
Foveal feedback uses the same representational format as bottom-up sensory drive in foveal V1.
- _tests_ **prediction-cross-decoding-generalizes** [prediction, hypothesis, epistemic=prediction, status=unknown]
The shared-format hypothesis predicts that classifiers trained on feedback responses should cross-decode to direct-stimulation responses.
- **foveal-feedback-below-direct-stimulation** [empirical, fig2A, epistemic=strong, status=unverified:compute-infeasible]
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.
- **foveal-v1-decodes-peripheral-saccade-target** [empirical, fig2A, epistemic=strong, status=unverified:compute-infeasible]
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.
- **lo-shows-reversed-specificity** [empirical, fig3B, epistemic=strong, status=unverified:compute-infeasible]
Lateral occipital cortex shows the inverse profile to foveal V1 — cross-shape decoding drops while cross-category decoding survives — yielding a double dissociation that rules out a generic sensitivity argument for the V1 result.
- _supports_ **hypothesis-feedback-carries-shape-not-category** [hypothesis, hypothesis, epistemic=hypothesis, status=unknown]
Foveal feedback to early visual cortex carries shape information but not semantic category — the level of representation appropriate to V1, not LO.
- _supports_ **decoding-shape-sensitive-not-semantic** [synthesis, fig3, epistemic=moderate, status=unverified:compute-infeasible]
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.
- _tests_ **prediction-lo-inverse-pattern** [prediction, hypothesis, epistemic=prediction, status=unknown]
The level-of-representation hypothesis predicts that LO will show the inverse profile to foveal V1 — cross-shape drops, cross-category preserved.
- _dissociates-with_ **v1-category-decoding-drops-in-feedback** [empirical, fig3B, epistemic=moderate, status=unverified:compute-infeasible]
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.
- **v1-category-decoding-drops-in-feedback** [empirical, fig3B, epistemic=moderate, status=unverified:compute-infeasible]
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.
- _supports_ **hypothesis-feedback-carries-shape-not-category** [hypothesis, hypothesis, epistemic=hypothesis, status=unknown]
Foveal feedback to early visual cortex carries shape information but not semantic category — the level of representation appropriate to V1, not LO.
- _supports_ **decoding-shape-sensitive-not-semantic** [synthesis, fig3, epistemic=moderate, status=unverified:compute-infeasible]
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.
- _tests_ **prediction-v1-category-drops-shape-preserved** [prediction, hypothesis, epistemic=prediction, status=unknown]
The shape-not-category hypothesis predicts that V1 cross-category decoding should drop under feedback while cross-shape decoding should be preserved.
- _dissociates-with_ **lo-shows-reversed-specificity** [empirical, fig3B, epistemic=strong, status=unverified:compute-infeasible]
Lateral occipital cortex shows the inverse profile to foveal V1 — cross-shape decoding drops while cross-category decoding survives — yielding a double dissociation that rules out a generic sensitivity argument for the V1 result.
- **v2-v3-generalize-shape-not-category** [empirical, fig3-figure-supplement-1, epistemic=moderate, status=unverified:compute-infeasible]
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.
- _supports_ **hypothesis-feedback-carries-shape-not-category** [hypothesis, hypothesis, epistemic=hypothesis, status=unknown]
Foveal feedback to early visual cortex carries shape information but not semantic category — the level of representation appropriate to V1, not LO.
- _tests_ **prediction-v1-category-drops-shape-preserved** [prediction, hypothesis, epistemic=prediction, status=unknown]
The shape-not-category hypothesis predicts that V1 cross-category decoding should drop under feedback while cross-shape decoding should be preserved.
▸ Show comparator prompt
You are comparing two articulations of the same paper's argument: (a) the published abstract, and (b) a synthesis reconstructed from the paper's enriched claim graph by an agent with no access to the abstract. The claim graph carries explicit edges representing argumentative forms — hypothesis (`entails`), prediction (`derived-from`), test (`tests`), abductive support (`supports`/`refutes`), elimination (`rules-out`), dissociation (`dissociates-with`), control validation (`validates`), interpretation (`interprets`), methodological warrant (`enables-method`), and scope qualification (`scopes`). Your job: identify what each surfaces, what each hides, and what the divergences tell us — particularly about the inferential structure of the argument and how it's compressed in the abstract. Look for: - Inferential structure surfaced in synthesis but flattened in abstract - Claims surfaced in synthesis but absent from abstract — categorize - Assertions in abstract but not in synthesis — categorize - Claims framed differently - Ordering and emphasis - Treatment of scope and caveats Output: 1. 200-word comparison paragraph 2. Structured table 3. Single most diagnostic divergence 4. Note on inferential structure recovery ABSTRACT: Human vision is characterized by frequent eye movements and constant shifts in visual input, yet our perception of the world remains remarkably stable. Here, we directly demonstrate image-specific foveal feedback to primary visual cortex in the context of saccadic eye movements. To this end, we used a gaze-contingent fMRI paradigm, in which peripheral saccade targets disappeared before they could be fixated. Despite no direct foveal stimulation, we were able to decode peripheral saccade targets from foveal retinotopic areas, demonstrating that image-specific feedback during saccade preparation may underlie this effect. Decoding was sensitive to shape but not semantic category of natural images, indicating feedback of only low-to-mid-level information. Cross-decoding to a control condition with foveal stimulus presentation indicates a shared representational format between foveal feedback and direct stimulation. Moreover, eccentricity-dependent analyses showed a U-shaped decoding curve, confirming that these results are not explained by spillover of peripheral activity or large receptive fields. Finally, fluctuations in foveal decodability covaried with activity in the intraparietal sulcus, thus providing a candidate region for driving foveal feedback. These findings suggest that foveal cortex predicts the features of incoming stimuli through feedback from higher cortical areas, which offers a candidate mechanism underlying stable perception. SYNTHESIS V3: 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) — 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 — 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.