## Moving the fixation point — a lesson in what the stimulus actually requires
2026 · Transparent Motion · G.S.
---
One of the cleaner manipulations available in a spatial attention paradigm is to move the fixation point. If an asymmetry in performance depends on where the observer is fixating relative to the stimulus — rather than on some intrinsic property of the stimulus geometry — then shifting fixation should shift or reverse the asymmetry. It is a standard control, conceptually clean, and requires no new stimulus code. We tried it.
The idea was to test whether the Near/Far performance asymmetry in the depth-separated version of the paradigm was fixation-centered — that is, driven by where attention naturally spreads relative to the observer's fixation plane — or whether it was intrinsic to the disparity structure of the stimulus regardless of where you were looking. Moving fixation laterally, away from the centre of the rotating dot fields, would dissociate the two.
What happened: the task became essentially impossible.
The transparent motion paradigm turns out to require central or near-central fixation to work at all. When fixation was shifted laterally, the two counter-rotating fields could no longer be perceptually segregated — they collapsed into a single incoherent motion signal. Direction judgments were near chance not because the cue was ineffective but because the translating field could not be identified in the first place. The stimulus simply did not support the percept from the periphery.
This is not entirely surprising in retrospect. Transparent motion perception is known to be fragile at high eccentricities, and the aperture used (Ap 3.5° radius, 7° diameter) places a lot of the dots at eccentricities where motion segregation is already taxed. Moving fixation away from centre pushed the entire stimulus further into the periphery, past the point where the visual system could maintain two distinct motion surfaces.
The practical lesson: fixation position is not a free parameter in this paradigm — it is load-bearing. The lateral shift experiment never made it into the dataset. But it made explicit a constraint that had been implicit in every other session: the stimulus is designed for foveal engagement, and that design assumption needs to be respected rather than tested.
[Note: details of shift magnitude and direction to be confirmed — draft entry for review.]