Skip to content

Predictive Remapping

Explain trans-saccadic visual stability by having visuospatial neurons shift their receptive fields to a stimulus's future post-saccadic retinal position before the eyes move — driven by an efference copy of the saccade command, so the eyes land on an already-built representation.

Core Idea

Predictive remapping is the neuroscientific mechanism by which neurons in visuospatial areas — principally lateral intraparietal cortex (LIP) and frontal eye fields (FEF) — shift their receptive fields to the future, post-saccadic retinal position of a stimulus before the saccade is executed, so that a representation of the target's expected post-movement location is already built by the time the eyes land. The defining structural commitments are: a retinotopic coordinate frame in which visual inputs are coded; an imminent saccade that will abruptly shift the retinal projection of the entire visual scene; an efference copy of the motor command to the eyes that is routed to sensory areas as a predictive signal; and a pre-saccadic receptive-field shift driven by that signal, in which neurons whose current receptive field covers the target's future retinal location begin responding to the target before the eyes move. The consequence is trans-saccadic perceptual continuity: the visual world does not appear to jump when the eyes saccade, despite the retinal image undergoing a large, rapid displacement, because the post-saccadic representation was partially pre-computed. A closely related corollary is that spatial attention is also pre-allocated to the saccade target before execution — the remapping does attentional work alongside representational work. The phenomenon was documented in awake, behaving monkeys by Duhamel, Colby, and Goldberg (1992), who recorded LIP neurons whose receptive fields shifted to their future location before the saccade, and has since been replicated and extended across parietal and frontal areas and in human neuroimaging.

Structural Signature

Sig role-phrases:

  • the retinotopic coordinate frame — the representational space (in LIP, FEF, superior colliculus) in which visual inputs are coded by retinal position
  • the imminent saccade — an about-to-execute eye movement that will abruptly displace the retinal projection of the entire scene
  • the efference copy — a corollary discharge of the saccade motor command, routed to sensory areas as the predictive signal that gates the shift
  • the pre-saccadic receptive-field shift — neurons whose field will cover the target's future retinal location beginning to respond to it before the eyes move
  • the future-location representation — the post-saccadic percept partly pre-computed, so the eyes land on an already-built representation
  • the trans-saccadic continuity — the visual world not appearing to jump despite the large retinal displacement, with no post-hoc reconciliation gap
  • the attentional pre-allocation — the same pre-saccadic signal also sending spatial attention to the saccade target ahead of the eyes
  • the saccade-vector scaling — the magnitude and direction of the shift matching the impending saccade, and gated strictly to the pre-saccadic window

What It Is Not

  • Not post-saccadic reconciliation. The mechanism's whole content is that the receptive-field shift leads the saccade: the post-saccadic location is partly pre-computed from an efference copy before the eyes move, so there is no after-the-fact matching of old retinal image to new. A response that depends on the stimulus having already entered the field is an ordinary visual response, not remapping.
  • Not an ordinary feedforward visual response. The defining observation is a neuron responding to a stimulus that is not yet in its receptive field but will be after the eyes land — so the response cannot be driven by current retinal input. It is driven by a predictive (corollary-discharge) signal, which is what makes it remapping rather than feedforward vision.
  • Not learning or durable recalibration. Predictive remapping is a one-shot, online update conditioned on the impending motor command, bound to the pre-saccadic window. A change that persists as an experience-driven recalibration across trials is learning, a different mechanism — remapping resets every saccade rather than accumulating.
  • Not purely representational. The same pre-saccadic signal does two jobs: it builds the future-location representation and pre-allocates spatial attention to the saccade target ahead of the eyes. Reading every pre-saccadic enhancement as representational remapping ignores the attentional component the mechanism also drives; the two must be separated by whether the effect is spatially specific to the remapped field.
  • Not generic anticipatory state update. A camera-stabilization buffer pre-shift, UI predictive pre-fetch, or an ML attention prefetch instantiates the parents feedforward/predictive_coding/forward_model, not predictive remapping — they lack a retinotopic coordinate frame whose projection is about to be displaced, a saccade as the conditioning action, the receptive-field-shift implementation, and the efference-copy gating. Stripped of those, the named mechanism reduces to "update the model ahead of an expected change," which the parents already supply.

Scope of Application

Predictive remapping lives within visuospatial and action-perception neuroscience; its reach is bounded there — a retinotopic coordinate frame whose projection a saccade is about to displace, with an efference copy gating a receptive-field shift. The engineered look-alikes (camera-buffer pre-shift, UI predictive prefetch) are instances of its parents feedforward / predictive_coding / forward_model, not of the named mechanism, and stay out.

  • Visuospatial neuroscience — the origin and home: pre-saccadic receptive-field remapping in lateral intraparietal area (LIP), frontal eye fields (FEF), and superior colliculus, documented by Duhamel, Colby, and Goldberg (1992), underwriting trans-saccadic perceptual continuity.
  • Human cognitive neuroimaging — the same anticipatory-shift signature replicated and extended in human parietal and frontal areas, and tied to the pre-saccadic pre-allocation of spatial attention.
  • Action-perception neuroscience — the broader efference-copy and forward-model accounts of motor anticipation (Wolpert and Ghahramani), where the predicted sensory consequence of an action is used to handle self-generated input; predictive remapping is the saccadic special case within this family.

Clarity

Predictive remapping's clarifying force is that it dissolves the long-standing puzzle of trans-saccadic perceptual stability: why the visual world does not appear to jump even though every saccade slams the retinal image of the whole scene to a new position several times a second. Before the mechanism was identified, the natural assumption was that vision must reconcile after the fact — wait for the eyes to land, then match the new retinal image against the old. Predictive remapping reframes the problem by showing the representation is shifted before the eyes move: the post-saccadic location is partly pre-computed from an efference copy of the motor command, so there is no post-hoc gap to paper over. That converts "how does perception stay stable across saccades?" from a mystery about reconciliation into a concrete question about when and where receptive fields shift and what signal drives the shift.

The concept also sharpens distinctions that retinotopic coding alone leaves blurred. It separates the coordinate frame an input is currently coded in from the frame it will occupy after the movement, making "remapping to the future receptive-field location" a precise, recordable event rather than a vague claim that the brain compensates for eye movements. And it exposes that the same pre-saccadic signal does two jobs at once — building the future representation and pre-allocating spatial attention to the saccade target — which lets a researcher ask whether a given pre-saccadic response reflects representational remapping, attentional pre-allocation, or both. That, in turn, makes the driving signal itself the object of study: if remapping is gated by an efference copy of the saccade command, then disrupting that command should disrupt the shift, turning perceptual continuity into something experimentally manipulable rather than merely described.

Manages Complexity

Around every saccade a visuospatial neuroscientist confronts a cluster of separate-looking phenomena: the visual scene staying put despite the retinal image jumping several times a second, spatial attention arriving at the saccade target before the eyes do, the suppression of the smear the rapid retinal displacement should produce, and the post-saccadic image being read out without a perceptible reconciliation gap. Treated independently, each invites its own post-hoc story — a separate stabilization process, a separate attention-shifting process, a separate blur-gating process — and each leaves the central question (how the brain reconciles two retinal images after the eyes land) as a standing mystery. Predictive remapping collapses that constellation onto one pre-saccadic event: neurons whose receptive field will cover the target's post-saccadic retinal location begin responding to it before the saccade, driven by an efference copy of the motor command. Once that single shift is posited, perceptual continuity, attention pre-allocation, and the absence of a post-hoc gap all fall out of it rather than each requiring a mechanism of its own, and the explanatory burden drops from a list of trans-saccadic puzzles to one anticipatory operation.

The compression turns a vague "the brain compensates for eye movements" into a few recordable parameters the researcher tracks and reads outcomes off, exactly as the Clarity sets up. The coordinate-frame distinction — current retinal frame versus future post-saccadic frame — makes where the receptive field shifts a measurable quantity rather than a hand-wave; the timing of the shift relative to saccade onset is a second; and the driving signal, the efference copy of the saccade command, is the third, with the branch that gating it (by disrupting the command) should abolish the shift and with it the continuity. So instead of modeling the full retinotopic-to-percept transformation case by case, the analyst carries one pre-saccadic shift indexed by frame, timing, and motor-copy signal, and predicts from it whether a given pre-saccadic response is representational remapping, attentional pre-allocation, or both, and whether perception will remain stable — the move from a scatter of trans-saccadic results to a single anticipatory mechanism with a small, experimentally manipulable parameter set.

Abstract Reasoning

Predictive remapping licenses a focused set of inferences within visuospatial neuroscience, each grounded in the pre-saccadic, efference-copy-driven receptive-field shift.

Diagnostic — reason from the response to the mechanism. The signature observable is a neuron in LIP or FEF that begins responding to a stimulus before the saccade, when the stimulus currently lies outside the neuron's receptive field but will fall inside it after the eyes land. From that anticipatory response the researcher infers a receptive-field shift toward the future (post-saccadic) retinal location, driven by a predictive signal rather than by current retinal input — the response cannot be a feedforward visual response because the stimulus is not yet in the field. The framework then forces a finer diagnostic it makes decidable: a pre-saccadic response could reflect representational remapping (the future location being pre-built) or attentional pre-allocation (processing resources sent to the saccade target ahead of the eyes), or both, since the same signal does both jobs; the analyst separates them by whether the enhancement is spatially specific to the remapped receptive-field location and tied to the coordinate-frame shift (remapping) versus a general gain increase at the target (attention). At the perceptual level, the diagnostic runs from stability back to pre-computation: because the world does not appear to jump despite a large retinal displacement each saccade, infer the post-saccadic representation was partly built in advance, so there is no post-hoc reconciliation gap to explain — the absence of a perceived jump is evidence the shift preceded the movement.

Interventionist — reason from the driving signal to a manipulation with a predicted effect. Because the shift is gated by an efference copy of the saccade command routed to sensory areas, the intervention space targets that signal and carries directional predictions. Disrupt the motor command or its corollary discharge (lesion or inactivation of the relay pathway, or its mediodorsal-thalamic route): predicted to abolish or attenuate the pre-saccadic receptive-field shift, and with it the trans-saccadic continuity that the shift produces. Perturb the saccade-related signal with TMS over parietal or frontal eye-field cortex timed to the pre-saccadic interval: predicted to degrade remapping specifically when delivered in the window before saccade onset, not after. Vary saccade parameters: predicted to scale the magnitude and direction of the receptive-field shift with the vector of the impending saccade, since the shift must match where the retinal image is about to go. The interventionist move is therefore a reading of one variable: act on the efference copy or its timing and predict the shift weakens, turning perceptual continuity from something merely described into something experimentally manipulable; the degree of disruption indexes how much of the stability the pre-computation was supplying.

Boundary-drawing — reason about when the mechanism is in force. Predictive remapping operates only where there is a retinotopic coordinate frame whose projection is about to be displaced, an imminent saccade, and an efference copy of that motor command available to drive the shift; absent a planned saccade there is no predictive signal and no pre-saccadic shift — drift and slow pursuit, which lack the abrupt frame change and the discrete motor command, are outside the concept. The shift is also strictly anticipatory and online, bounded to the pre-saccadic window: a response that depends on the stimulus having already entered the field is an ordinary visual response, not remapping, and a change that persists as a durable experience-driven recalibration is learning, not the one-shot motor-conditioned update. The boundary inference runs: if a candidate response is present without an impending saccade, or appears only after the eyes have moved, the efference-copy-gated remapping mechanism is the wrong account.

Predictive / order-of-events. The mechanism fixes a strict sequence — saccade command issued → efference copy routed to visuospatial sensory areas as a predictive signal → neurons whose receptive field will cover the target's post-saccadic location begin responding to the target, before the eyes move → saccade executed, retinal image displaced → the eyes land on an already-partly-built representation, yielding continuity with no reconciliation gap. This ordering licenses predictions: that the receptive-field shift leads saccade onset rather than following it (the defining temporal claim, recordable as shift-time relative to movement); that attention arrives at the saccade target before the eyes do, because the same pre-saccadic signal pre-allocates it; and that the post-saccadic percept is read out without a perceptible smear or jump precisely because the future state was computed ahead of the movement rather than reconciled after it — so anything that delays the efference copy past saccade onset is predicted to restore the post-hoc gap the mechanism normally pre-empts.

Knowledge Transfer

Within visuospatial and action-perception neuroscience predictive remapping transfers as mechanism, with the efference-copy-gated, pre-saccadic receptive-field shift as the portable core. The mechanism documented in LIP (Duhamel, Colby, and Goldberg) generalizes across the parietal and frontal eye fields and the superior colliculus, and into human neuroimaging, and it sits in the same family as the broader efference-copy/forward-model accounts of motor anticipation (Wolpert and Ghahramani), where a predicted sensory consequence of action is used to handle self-generated input. Across these the diagnostics and interventions carry intact: an anticipatory response to a stimulus not yet in the receptive field implies a shift toward the future retinal location; disrupting the corollary discharge (lesion or inactivation of the relay, or TMS timed to the pre-saccadic window) is predicted to abolish the shift and the continuity; the shift's magnitude and direction scale with the saccade vector. The coordinate-frame distinction (current versus post-saccadic retinal frame), the timing-relative-to-saccade, and the motor-copy signal are the three recordable parameters, and the same separation of representational remapping from attentional pre-allocation applies wherever the mechanism is studied. The within-domain transfer is the anticipatory-shift mechanism itself moving across visuospatial areas and into the motor-anticipation literature.

Beyond the visuospatial saccade case the picture is the third category, and the parents are precisely identified. The genuinely substrate-independent residue is update an internal model in advance of an expected state change — anticipatory state update — and that pattern is already carried by the parent primes feedforward (act before the disturbance), predictive_coding (the system predicts its own input), and the forward_model pattern (a system predicts the sensory consequences of its own action). Those parents travel, and the cross-domain look-alikes are genuine instances of them: a camera-stabilization system pre-shifting its image buffer in anticipation of a pan, UI predictive rendering that pre-fetches the next likely view, a machine-learning attention prefetch, an organization pre-positioning roles before a reorganization. But these are engineered instances of feedforward/forward-modeling, not of predictive remapping, because they lack its distinctive cargo: a retinotopic coordinate frame whose projection is about to be displaced, a saccade as the specific conditioning action, the receptive-field-shift implementation, and the efference-copy gating. Stripped of those, predictive remapping reduces to "anticipatory state update," which the parents already supply. So calling a camera's buffer pre-shift "predictive remapping" would import the saccade-and-retinotopy specifics that have no referent in a stabilizer — analogy at the level of "compute the future state in advance to avoid post-action reconciliation," mechanism only at the level of feedforward/predictive_coding/forward_model. The honest move is therefore layered: within neuroscience the efference-copy-gated receptive-field-shift mechanism and its three-parameter diagnostics travel across visuospatial areas and the motor-anticipation literature; the abstract "update the model ahead of an expected change" lesson belongs to feedforward, predictive_coding, and the forward_model pattern wherever an engineered or biological system anticipates a state change; but "predictive remapping," as named — retinotopic frame, saccade-conditioned, receptive-field shift — is reserved for the visuospatial neuroscience case (see Structural Core vs. Domain Accent).

Examples

Canonical

The founding demonstration is Duhamel, Colby, and Goldberg (1992), recording single neurons in the lateral intraparietal area (LIP) of awake monkeys performing saccades. They found neurons that began responding to a stimulus placed at the location their receptive field would occupy only after an impending saccade — before the eyes had moved and while the stimulus still lay outside the neuron's current receptive field. The anticipatory response could not be an ordinary visual response, since no light from the stimulus was yet falling in the field; it appeared only when a saccade was planned that would bring that location into the field. The receptive field had, in effect, shifted to its future retinal position ahead of the movement.

Mapped back: The LIP map is the retinotopic coordinate frame; the planned eye movement is the imminent saccade, and its efference copy is what lets the neuron respond before any retinal input arrives. The neuron firing to the stimulus's post-saccadic location before the eyes move is the pre-saccadic receptive-field shift building the future-location representation. That the shift is contingent on a planned saccade — absent one, no anticipatory response — demonstrates the saccade-vector scaling and gating.

Applied / In Practice

The efference-copy signal that gates remapping is the object of a clinical-research program in schizophrenia. On the forward-model account, several first-rank symptoms — the sense that one's own thoughts or actions are externally controlled — reflect a faulty corollary discharge: the brain fails to tag self-generated events with an accurate prediction of their sensory consequences. Because trans-saccadic stability depends on exactly this efference-copy signal, oculomotor paradigms that probe pre-saccadic updating (for example, judging the stability of a target displaced during a saccade, or double-step saccade tasks) are used as measurable assays of corollary-discharge integrity, and patients show reduced trans-saccadic updating consistent with a weakened predictive signal.

Mapped back: The paradigm isolates the efference copy/corollary discharge that normally drives the pre-saccadic receptive-field shift. A weakened signal degrades the future-location representation and hence trans-saccadic continuity, so measuring stability judgments across the imminent saccade turns the internal predictive signal into a clinical readout — the interventionist logic of acting on (or assaying) the driving signal rather than merely describing the percept.

Structural Tensions

T1: Pre-computation versus prediction error (anticipation buys continuity by committing before the eyes land). The mechanism's whole payoff is that it builds the post-saccadic representation ahead of the movement from an efference copy, so the eyes land on an already-built percept with no reconciliation gap. But committing in advance is committing to a prediction, and the prediction can be wrong: if the target moves during the saccade, or the saccade lands off its intended vector, the pre-computed representation mismatches the actual post-saccadic retinal input, and the system must still absorb the error (the basis of intrasaccadic mislocalization and saccadic suppression of displacement — where sizable target jumps go unnoticed exactly because the prediction, not the input, is trusted). So predictive remapping does not eliminate reconciliation; it front-loads a bet and pays when the bet is wrong. The very anticipation that removes the post-hoc gap creates a prediction-error regime whenever the world changes mid-saccade. Diagnostic: Did the world stay put across the saccade (pre-computation succeeds silently), or did it change intrasaccadically (the trusted prediction now produces mislocalization the mechanism must absorb)?

T2: Representational remapping versus attentional pre-allocation (one signal doing two jobs threatens the concept's autonomy). The entry stresses that the same pre-saccadic efference-copy signal both builds the future-location representation and pre-allocates spatial attention to the saccade target. That dual role is a genuine tension for the concept's identity: if a pre-saccadic enhancement can be a general gain increase at the attended target rather than a coordinate-frame-specific receptive-field shift, then "remapping" risks reducing to "presaccadic attention shifting," which is a live interpretation in the field. The concept's claim to be a representational mechanism distinct from attention depends on showing spatial specificity to the remapped field — but the very signal that would establish remapping is the one that also moves attention, so the two are confounded at their source. The autonomy of remapping-as-representation from attention-as-gain is exactly what the shared driver makes hard to secure. Diagnostic: Is the pre-saccadic enhancement spatially specific to the future receptive-field location and tied to the frame shift (remapping), or a general gain increase at the saccade target (attention) — and can they even be dissociated when one signal drives both?

T3: One anticipatory shift versus the several mechanisms of trans-saccadic stability (elegant compression that may over-attribute). The concept's power is compressing continuity, attention pre-allocation, blur suppression, and the absent reconciliation gap onto a single pre-saccadic shift. But that unification may claim too much: remapping is selective, shifting a sparse set of behaviorally relevant receptive fields, not the whole retinotopic map, so it does not obviously deliver global scene stability by itself; and saccadic suppression of the intra-saccadic smear is plausibly a separate gating mechanism, not a corollary of the shift. Reading every trans-saccadic phenomenon as "falling out of" one anticipatory operation risks folding distinct processes into the flagship mechanism. The explanatory economy that makes the concept attractive is also a pressure to attribute the entire stability problem to remapping when the phenomenon may require a coalition of mechanisms. Diagnostic: Does the phenomenon in question follow from the selective receptive-field shift itself, or is it a distinct process (whole-scene stability, saccadic suppression) being absorbed into remapping for tidiness?

T4: Manipulable driving signal versus the confound of disrupting it (a clean test that perturbs more than remapping). Framing the shift as gated by an efference copy makes it experimentally manipulable — disrupt the corollary discharge and the shift, and the continuity, should weaken, which is the concept's great methodological asset. But the efference copy is upstream of the saccade itself and of attentional pre-allocation, so lesioning the relay, inactivating the thalamic route, or mistimed TMS degrades the movement and the attention shift too, not remapping alone. A degraded stability judgment after perturbation is therefore hard to attribute uniquely to lost remapping rather than to a disturbed saccade or misallocated attention. The intervention that turns perceptual continuity from description into manipulation is confounded precisely because the driving signal is shared. The testability the concept prizes and the specificity it needs pull against each other. Diagnostic: Does the disruption's effect isolate the receptive-field shift, or is the observed instability equally explained by a perturbed saccade or attention downstream of the same efference copy?

T5: Autonomy versus reduction (a named neuroscience mechanism or an instance of anticipatory forward modeling). Predictive remapping is a fully specified visuospatial-neuroscience mechanism with irreducibly local cargo — a retinotopic coordinate frame about to be displaced, the saccade as the conditioning action, the receptive-field-shift implementation, and efference-copy gating — and within neuroscience it transfers as mechanism across LIP, FEF, the superior colliculus, human neuroimaging, and the broader motor-anticipation literature. But beyond the saccadic case it does not travel as the named mechanism: the substrate-independent residue is update an internal model in advance of an expected state change, carried by feedforward, predictive_coding, and the forward_model pattern. A camera stabilizer pre-shifting its buffer or a UI prefetching the next view instantiates those parents, not predictive remapping — no retinotopy, no saccade, no receptive field. The tension is between a mechanism that earns its own oculomotor detail and the recognition that its cross-domain lesson belongs to the feedforward/forward-model parents. Diagnostic: Resolve toward feedforward / predictive_coding / forward_model when an engineered or biological system anticipates a state change generally; toward named predictive remapping when a saccade is about to displace a retinotopic frame and an efference copy gates a receptive-field shift.

Structural–Framed Character

Predictive remapping sits toward the structural pole — best read as mixed-structural, closely analogous to the precedence effect: a genuine, evaluatively-neutral neural mechanism running observer-free, held short of the pole by domain-pinned vocabulary. Four criteria run structural. Its evaluative weight is nil: a receptive field shifting to its future retinal location ahead of a saccade is a processing operation, neither good nor bad — no verdict. It is not human-practice-bound: the mechanism runs in the visuospatial cortex of awake behaving monkeys and humans on every saccade, several times a second, whether or not anyone records it — the efference-copy-gated shift is a fact of the circuit, not the output of a practice. Its institutional origin is none: it is a discovered neural mechanism (Duhamel, Colby, Goldberg 1992), named rather than invented. And within visuospatial and action-perception neuroscience cross-domain reuse is recognition of the same mechanism — it generalizes across LIP, FEF, and the superior colliculus, into human neuroimaging, and sits in the same efference-copy/forward-model family as broader motor anticipation, as literal mechanism.

What keeps it off the structural pole is vocab_travels, which it fails, with the import_vs_recognize flip beyond the saccade case: the operative vocabulary — retinotopic coordinate frame, saccade, efference copy/corollary discharge, pre-saccadic receptive-field shift — is bound to visuospatial neuroscience and does not float free, and engineered look-alikes (camera-buffer pre-shift, UI prefetch, ML attention prefetch) are instances of the parents, not of predictive remapping. The portable structural skeleton is anticipatory forward modeling — update an internal model in advance of an expected state change — carried by feedforward (act before the disturbance), predictive_coding (the system predicts its own input), and the forward_model pattern (predict the sensory consequences of one's own action). That skeleton is exactly what predictive remapping instantiates (the saccadic special case), not what makes "predictive remapping" travel: the cross-domain reach belongs to those parents, while the retinotopic frame, the saccade conditioning, the receptive-field-shift implementation, and the efference-copy gating are the domain accent that stays home. Its character: a real, evaluatively-neutral visuospatial-neural mechanism running observer-free every saccade, structural in the anticipatory forward-model skeleton (feedforward/predictive coding/forward model) it instantiates but pinned by retinotopy/saccade/efference-copy vocabulary to visuospatial neuroscience — mixed-structural, close to but short of the pole.

Structural Core vs. Domain Accent

This section decides why predictive remapping is a domain-specific abstraction and not a prime — marking where the portable anticipatory-forward-model skeleton ends and the visuospatial-neuroscience machinery begins.

What is skeletal (could lift toward a cross-domain prime). Strip the visuospatial neuroscience and a thin relational structure survives: a system with an internal representation, facing a self-generated action that will predictably displace its input, uses a copy of the action command to update the representation to its post-action state before the action executes — so it lands on an already-built representation with no after-the-fact reconciliation. The portable pieces are abstract — an internal model, a command copy that predicts the sensory consequence, and a pre-emptive update ahead of the change. This is feedforward (act before the disturbance), predictive_coding (the system predicts its own input), and the forward_model pattern (predict the sensory consequences of one's own action). The skeleton is genuinely substrate-portable — "update the model ahead of an expected change" recurs across engineered and biological systems — which is why the entry names those three parents as what actually carries any cross-domain lesson. That portable core is what predictive remapping instantiates (the saccadic special case), not what makes it predictive remapping.

What is domain-bound. Almost everything that makes the construct this mechanism is visuospatial-neuroscience furniture and none of it survives extraction: the retinotopic coordinate frame (in LIP, FEF, superior colliculus) whose projection is about to be displaced; the saccade as the specific conditioning action; the efference copy / corollary discharge of the saccade command as the gating signal; the pre-saccadic receptive-field shift as the implementation; the saccade-vector scaling and pre-saccadic-window gating; and the coupled attentional pre-allocation the same signal drives. These are the worked vocabulary, the instruments (single-unit recording, TMS timed to the pre-saccadic window, corollary-discharge assays), and the empirical cases (Duhamel-Colby-Goldberg; schizophrenia forward-model deficits) the field studies. The decisive test: a camera-stabilization buffer pre-shift, a UI predictive prefetch, or an ML attention prefetch has no retinotopy, no saccade, and no receptive field — so calling it "predictive remapping" imports saccade-and-retinotopy specifics with no referent; stripped of those, the named mechanism reduces to "anticipatory state update," which the parents already supply.

Why this does not clear the prime bar. A prime is a relational structure whose vocabulary travels and whose cross-domain transfer is recognition of the same mechanism, not analogy. Predictive remapping's transfer is bimodal. Within visuospatial and action-perception neuroscience it travels as full mechanism — the efference-copy-gated pre-saccadic receptive-field shift, its three recordable parameters (frame, timing, motor-copy signal), and the representational-versus-attentional diagnostic carry across LIP, FEF, the superior colliculus, human neuroimaging, and the broader efference-copy/forward-model motor-anticipation literature, of which it is the saccadic special case: genuine recognition of one mechanism. Beyond the saccade case it does not travel as predictive remapping: camera stabilizers, UI prefetch, and organizational pre-positioning are engineered instances of the parents feedforward/predictive_coding/forward_model, not of predictive remapping, and to name them so would be analogy at the level of "compute the future state in advance." And when the bare structural lesson is needed cross-domain — update an internal model ahead of an expected state change — it is already carried, in more general form, by those three parents. The cross-domain reach belongs to them; "predictive remapping," as named, carries retinotopy, saccade-conditioning, receptive-field-shift, and efference-copy machinery that stays bound to visuospatial neuroscience.

Relationships to Other Abstractions

Local relationship map for Predictive RemappingParents appear above the current abstraction, mutual partners to the right, and children below. Node labels state whether each abstraction is prime or domain-specific; colors identify relation types.Predictive RemappingDOMAINPrime abstraction: Feedforward — is a kind ofFeedforwardPRIME

Current abstraction Predictive Remapping Domain-specific

Parents (1) — more general patterns this builds on

  • Predictive Remapping is a kind of Feedforward Prime

    Predictive remapping is feedforward compensation specialized to using an efference copy to shift a visual representation before an impending saccade changes its input coordinates.

Hierarchy path (1) — routes to 1 parentless root

Not to Be Confused With

  • Efference copy / corollary discharge. The driving signal — a copy of the saccade motor command routed to sensory areas. Predictive remapping is the receptive-field shift that this signal gates, not the signal itself; the efference copy is the cause/input, remapping is the effect/implementation. The same corollary discharge also does other jobs (attentional pre-allocation, saccadic suppression). Cause-versus-mechanism. Tell: is the referent the motor-command copy sent to sensory cortex (efference copy), or the anticipatory receptive-field shift it triggers (predictive remapping)?

  • Presaccadic attention shift. The pre-allocation of spatial attention (a gain increase) to the saccade target before the eyes move — driven by the same pre-saccadic signal, which is exactly why it is confounded with remapping. But it is a general enhancement at the target, whereas remapping is a coordinate-frame-specific shift to the future receptive-field location. Sibling process sharing a driver, distinguished by spatial specificity. Tell: is the pre-saccadic enhancement a general gain boost at the saccade target (attention shift), or specific to the remapped future field location (predictive remapping)?

  • Saccadic suppression. The reduction of visual sensitivity during the saccade that blanks the intra-saccadic motion smear. It solves a different piece of trans-saccadic stability (suppressing the blur) and is plausibly a separate gating mechanism, not a corollary of the receptive-field shift. Coalition-member, not synonym; folding it into remapping over-attributes. Tell: is the phenomenon the blanking of sensitivity during the eye movement (saccadic suppression), or the pre-saccadic shift of receptive fields to future locations (predictive remapping)?

  • Spatial updating / trans-saccadic constancy. The broader phenomenon of perceiving a stable world across eye movements (the explanandum). Predictive remapping is one proposed mechanism for it — but a selective one (shifting sparse behaviorally-relevant fields), not obviously the whole story, which may need saccadic suppression and other processes. Phenomenon-versus-its-candidate-mechanism, part-versus-whole. Tell: is the referent the overall achievement of perceptual stability across saccades (spatial constancy), or the specific pre-saccadic field-shift posited to help produce it (predictive remapping)?

  • Feedforward / predictive coding / forward model (parent primes). The substrate-neutral core predictive remapping instantiates — updating an internal model ahead of an expected, self-generated state change. These carry to camera-buffer pre-shifts, UI prefetch, and ML attention prefetch; predictive remapping is their saccadic special case with retinotopy and efference-copy gating. Treated more fully in the Knowledge Transfer and Structural Core vs. Domain Accent sections. Tell: strip the retinotopic frame and saccade conditioning and what remains — anticipatory model update ahead of a predicted change — is the parent, not predictive remapping.

Neighborhood in Abstraction Space

Predictive Remapping sits in a sparse region of the domain-specific corpus (66th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Neural Topographic Maps (7 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-07-12