Do network expressions track transitions between fixation and stimulation?
The shared model is fitted without task labels. Aligning its expressions to stimulus onset and offset lets us examine whether a recurring pattern accompanies the experimental transitions. The condition-comparison page tests the complementary descriptive question: whether each participant’s average shifts in the same direction.
Explore the figures
Choose the full cohort to see the participant-balanced trajectory, or select a participant and run to inspect the recordings underneath it. Zero marks the stimulus boundary; shading shows variation, not certainty about a neural response. Use the boundary-delay comparison to judge how dependent the average difference is on timing.
How do shared networks change through visual stimulation?
A single rank-7 model describes all 60 separate runs from 15 participants. The question is whether the same learned spatial patterns change around task onset and offset across people, and how much individual recordings depart from that average. Task labels were not supplied during fitting.
Main findings for this selection
Exact selected recordings
Expression at task onset and offset
Which networks contribute most strongly?
Contribution magnitude is |expression| multiplied by the norm of that network’s unique off-diagonal coefficient weights. Here it is averaged over time, then runs and participants. Points show participant means; error bars show participant SD. This couples expression amplitude with matrix scale, but it is not explained BOLD variance, a causal effect or a probability. Network 1 is omitted from figures as a mean/bias-like component and retained in the model.
Does signed expression differ between task and fixation?
Task and fixation are compared within each network in original expression units. Increasing signed expression scales the matrix’s two coefficient poles together; negative expression reverses their contribution. The sign does not establish excitation or inhibition. Inter-block fixation contains delayed task responses and is not a separate resting scan.
How much does the contrast depend on timing?
Inspect one continuous recording
Scope of interpretation
Participant-balanced averages describe this cohort in the fitted representation. Shared spatial coordinates enable comparison but can constrain individual effects. Individual curves and participant SD help distinguish a group tendency from heterogeneous responses; they are not significance tests. Motion- and coverage-flagged recordings remain included. In particular, inspect V015’s motion before attributing its expression to task physiology.
Model, exact membership and summary provenance
More about the measures and interpretation
Participant controls change expression summaries, not the fixed shared coefficient maps in the anatomical tabs. These are descriptive in-sample associations, not activation probabilities or independently validated effects. Network numbers refer to this cohort fit, not the preserved single-participant fit.
Each expression was standardized over its full run. Blocks are averaged within run, runs within participant, then participants equally. Shading is ±1 SD across the stated units, not a confidence interval. Zero is the measured stimulus boundary; no HRF correction is applied. Three history samples span about four seconds. Neighbouring task blocks can enter the later parts of these overlapping windows.
The 4/6 s choices shift the entire block envelope and alter which samples enter task and fixation averages. They do not estimate an HRF, change the fit, or justify choosing the largest contrast. Initial and final fixation are excluded. Raw coordinate amplitudes are not common strength units across networks.
Chronological traces are shown only for individual recordings. Scans are never concatenated across run boundaries or averaged on an assumed common full-run stimulus schedule. Gold segments mark measured visual-stimulation blocks.