Evidence Tracing

Honest introspection: my mechanistic bias is well-matched to this design, but it has named failure modes

What the bias buys here. The design's logic — fit MCD taus pre/post and let the data adjudicate low-level tuning change vs readout re-weighting — only works if tau estimates track something physiological. Direct precedent exists: presynaptic silencing at the calyx of Held shifted postsynaptic membrane time constants by ~5 days, with Kv conductance changes reproducing the phenotype in single-compartment modeling (Calyx TeNT tau evidence). Intrinsic exponential taus (τ ~1–49 s) are robustly fittable and pharmacology-sensitive in CA1 (CA1 tau fitting methods). Sweep-SSVEP supplies the human-side measurement scaffolding for 1–20+ Hz dense frequency mapping (Sweep-SSVEP protocol template).

Where the bias could mislead me. (1) Measurement gap: system-level taus fitted from scalp EEG are not membrane taus; without an invasive or pharmacologically-anchored validation arm, a tau shift could reflect gain changes, not kinetics. (2) Both-lever confound: most synaptic-time-constant manipulations also change gain and arousal, so tau and readout coefficients may co-shift; mediation analysis and a neutral stimulus-control condition are mandatory, not optional. (3) Confirmation risk: because I prefer mechanistic stories, I would be tempted to accept a marginal tau shift as decisive; I should pre-register a tau-shift effect size and a falsification rule (stable taus + changed readout weights ⇒ readout-only). (4) Pure-behavioral designs are not worthless: for some arousal effects, readout-only models may be true, and behavioral-only paradigms test that efficiently; my preference should not dismiss them as inferior by default.

Decision structure the data will impose. The figure below is an illustrative schematic of the two outcome signatures, not measured data: a genuine tau shift moves the sensitivity curve shape across frequencies, whereas readout-only change leaves the curve fixed while classifier weights change. The key discriminant is curve shape change vs weight change at fixed shape.

What would change my conclusion: an independent physiological measure (e.g., slice or in vivo membrane-tau recording) under the same manipulation failing to co-vary with fitted MCD taus would show my EEG tau proxy is invalid; conversely, pre-registered mediation showing tau change explains the behavioral shift would confirm the mechanistic priority is warranted. Presynaptic calcium-current/release kinetics at fast synapses also demonstrably set frequency-following capacity (Calyx release kinetics), reinforcing that a well-chosen pharmacological lever should, in principle, shift tunable taus — the empirical question is whether the human readout resolves it.