Module 11 — Methods: how we actually know any of this

doo-wop classical, dance drill · 3:46

Listen on 93

Lyrics

[Verse 1]
Single units firing in the cortex maze
Neuropixels catching spikes through voltage waves
Two-photon calcium glowing bright and slow
Each method tells a different tale, you need to know
Electrophysiology fast but spatially blind
Calcium imaging sees cells but lags behind
Space and time trade-offs in every tool we choose
Pick your poison, accept what you will lose

[Chorus]
How do we know what neurons really do?
Correlation ain't causation, that's the truth
Decoders show what's there, not what gets used
Methods make the message, don't get confused
Record, stimulate, analyze, decode
Every claim needs evidence on that road

[Verse 2]
fMRI blood flow painting pictures slow
Spatial precision but temporal let-go
MEG and EEG racing at light speed
But where's it happening? Location's hard to read
ECoG electrodes sitting on the brain
Rare opportunity in the surgical game
Human methods limited by what we can do
Ethics and access constrain our view

[Chorus]
How do we know what neurons really do?
Correlation ain't causation, that's the truth
Decoders show what's there, not what gets used
Methods make the message, don't get confused
Record, stimulate, analyze, decode
Every claim needs evidence on that road

[Bridge]
Optogenetics flashing blue light beams
Chemogenetics and microstim schemes
Lesions show us when the system breaks
Causation finally for the research stakes
But Jonas and Kording dropped that mic
Even microprocessors fool our sight

[Verse 3]
Encoding models predict what cells should say
Decoding backwards works the other way
RSA comparing patterns in the space
Dimensionality reduction finds the trace
Deep networks mirror biological design
But models are hypotheses by design
Computational theories need the test
Artificial systems put us to the test

[Verse 4]
Cross-validation splits your data clean
Overfitting hides what patterns really mean
Bootstrap confidence around each claim
Reproducibility should be the aim
Population vectors point the neural way
But noise correlations change the play
Theory-driven questions guide the path
Not just fishing expeditions in the math

[Chorus]
How do we know what neurons really do?
Correlation ain't causation, that's the truth
Decoders show what's there, not what gets used
Methods make the message, don't get confused
Record, stimulate, analyze, decode
Every claim needs evidence on that road

[Outro]
Stay humble with the data that you find
Method-dependent truth should blow your mind
Calibrate each claim against its source
That's the neuroscientist's honest course

← Module 10 — Dynamics: computation through time | Capstone — Synthesis and the open frontier →