mne-erp

MNE ERP / ERF Analysis (grill → analyze → critic)

Evoked-response analysis of neurophysiology data via the MNE MCP server. This skill is skeptical by design: most ERP mistakes (peak-picking a window after seeing the grand average, a high-pass that shifts latency, calling an occipital 100 ms deflection "N1") run without any error — so the discipline is to grill the component identity and measurement plan before averaging, and critique before believing.

Companion skills: mne-mcp-guard for technical execution safety; mne-methodology-critic for Phase 3. Loaded objects persist in one MNE session.

---

PHASE 1 — GRILL (before averaging anything)

Do not average until these are answered. If the user can't answer one, propose a sensible default and explicitly flag the open risk — never silently choose.

Design & claim

• What is the hypothesis, and what is the comparison? (condition × condition, group × group, pre × post)
• Within- or between-subject? Paired or independent? n per cell?
• Confirmatory (component + window + electrodes pre-specified) or exploratory (whole-head, corrected)?

The questions that decide validity

• Which components, and what are their EXPECTED latency, topography, AND stimulus modality? Latency/polarity/topography must match the named component for that modality. ⚠️ auditory N1 is fronto-central (~100 ms); visual N1 is occipito-temporal ~150–200 ms — a ~100 ms occipital deflection in vision is most likely P1, not N1. A topography/modality mismatch means the component is misnamed. (This is the single most common identity error here.)
• Is the measurement window + electrode set PRE-SPECIFIED? A mean/peak amplitude measured in a window/channel chosen after seeing this dataset's grand average is circular (double-dipping). Pin the window and ROI down NOW, from prior literature, not from the data.

Data & parameters

• Baseline window (length + placement)? Baseline noise propagates into every component.
• Filter. ⚠️ high-pass > 0.3 Hz distorts slow components and shifts/inverts apparent latency and polarity (CDA, LPP, P300, CNV especially); low-pass smooths and delays peaks. State the band.
• Epoching: tmin/tmax, peak-to-peak rejection threshold + its justification.
• Reference, montage, channel selection, units.