Biofeedback is the process of measuring physiological signals from the body, processing them in real time, and presenting an output — a sound, a number, a display — that reflects some feature of that signal, enabling a person to attempt to modulate it. This layer maps the biofeedback process onto the Primordial Brain Layer v0.1 framework.
The central architectural claim of this layer, and the claim that requires the most careful HIR governance: biofeedback signals are distal proxies, not direct measurements of specific brain structures. A scalp EEG electrode is separated from the brain region it is most commonly associated with by volume conduction through scalp, skull, meninges, and cerebrospinal fluid, plus the mixing of signals from large cortical populations. An HRV measurement is separated from the autonomic control structures of the brainstem and hypothalamus by the entire peripheral autonomic chain. The feedback loop is real and measurable; the causal path from feedback display back to a specific brain region is probabilistic and requires separate validation at each step.
| In scope | Out of scope |
|---|---|
| Physiological signal modalities and what they directly measure | Biofeedback as a proven mechanism for reorganizing specific brain structures |
| The feedback loop architecture as an engineered system | Neurofeedback as a diagnostic tool |
| Protocol definitions with evidence classifications | Biofeedback signals as objective measures of mental states |
| Training level taxonomy (session state vs. skill vs. generalization vs. mechanism) | Biofeedback as a replacement for clinical evaluation or treatment |
| Signal-to-brain-layer mapping with explicit uncertainty | Overclaims about specific brain regions being "trained" without mechanism evidence |
| Extended uncertainty taxonomy for biofeedback-specific unknowns | Character, identity, or diagnostic inferences from biofeedback signals |
| HIR rules R-13 through R-22 extending the brain layer rule set | Any claim not grounded in cited, bounded evidence |
The biofeedback loop is an engineered system with seven stages. HIR governance applies at each stage independently — a clean signal at stage 2 does not compensate for a provenance failure at stage 1, and a valid feedback display at stage 5 does not authorize mechanistic claims about stage 7. Each gate is independent.
| Loop Stage | What it produces | What it does NOT establish | Key uncertainty classes |
|---|---|---|---|
| FL-1 — Sensor | Raw physiological voltage or current | Specific brain region activity | BB-01 BB-02 |
| FL-2 — Acquisition | Digitized signal with known sampling parameters | Clean, artifact-free signal (must be verified) | UB-04 BB-03 |
| FL-3 — Preprocessing | Filtered, referenced, artifact-rejected signal | Ground truth; preprocessing choices alter outcomes | UB-12 BB-04 |
| FL-4 — Feature | A scalar metric (band power, HRV, RMS) | The brain state the metric is "associated with" | BA-01 BA-04 |
| FL-5 — Feedback | A display signal that responds to the feature | Clinical outcome, psychological state, diagnosis | BA-02 BA-05 |
| FL-6 — Response | A behavioral attempt to modulate the signal | Which cognitive strategy worked; why | BA-03 BA-06 |
| FL-7 — Change | Within-session signal shift | Lasting neural change, generalization, or mechanism | BA-01 BA-07 |
For each signal, the most important entry is what it directly measures — not what it is commonly claimed to measure. The gap between these two is where most overclaiming occurs.
The most important governance concept in this layer. Every biofeedback outcome claim must be classified by its training level. Demonstrating Level 1 does not validate any higher level. Most overclaiming in biofeedback occurs when Level 1 evidence is used to assert Level 3 or Level 4 outcomes. Each level requires its own separate validation.
| Protocol ID | Name | Signal | Target | Brain Layer Links | Evidence Level | Training Level Evidence | Overclaim Risk |
|---|---|---|---|---|---|---|---|
| BP-01 | SMR / Theta-Beta Neurofeedback |
BF-01 (EEG) SMR ↑ (12–15 Hz) at C3/Cz/C4 Theta ↓ at Fz |
Attention · impulse regulation | BR-05 (indirect), OUT-10, OUT-11 | Level B | Level 1–2 demonstrated; Level 3 contested; Level 4 not established | High — "trains the ADHD brain" not established; mechanism unconfirmed |
| BP-02 | Alpha Enhancement Neurofeedback |
BF-01 (EEG) Alpha ↑ (8–12 Hz) Typically Pz or Oz |
Relaxation · anxiety reduction | BR-07 (visual alpha suppression), BR-01 (frontal alpha), OUT-12 | Level C | Level 1 demonstrated; Level 3 insufficient; Level 4 absent | Medium — "creativity enhancement" claims not established; relaxation ≠ alpha alone |
| BP-03 | Slow Cortical Potential Neurofeedback (SCP) |
BF-01 (EEG) Slow cortical potential direction (positive/negative shifts) |
Cortical excitability · ADHD · epilepsy | BR-01 (excitability), OUT-10, OUT-11 | Level B | Level 1–2 demonstrated; Level 3 evidence for ADHD and epilepsy exists but limited | Medium — Better mechanistic model than frequency-based; still requires careful clinical context |
| BP-04 | Frontal Alpha Asymmetry Neurofeedback |
BF-01 (EEG) F4 alpha − F3 alpha asymmetry |
Emotional regulation · approach motivation | BR-01 (prefrontal), OUT-12 | Insufficient | Based on group-level EEG emotion research; individual-level application not validated | High — "Training emotional states" claim not established; R-08 applies (group finding ≠ individual proof) |
| BP-05 | LORETA Neurofeedback | BF-01 (EEG) Source-localized current density estimate |
Region-specific neurofeedback | Any BR target (source localization dependent); adds UA-04 to all inherited uncertainty | Very limited | Source localization adds inverse problem uncertainty; Level 4 claim more pronounced but evidence weaker | High — "Directly targeting hippocampus" or "directly training amygdala" overclaims source localization certainty substantially |
| BP-06 | HRV Resonance Biofeedback |
BF-03 (HRV) Resonant frequency breathing (~0.1 Hz) |
Autonomic regulation · stress · anxiety | PW-07, BR-11, BR-13, OUT-12, OUT-13 | Level B | Level 2–3 evidence for anxiety, hypertension; mechanism partially understood (baroreflex, RSA) | Medium — Cardiac signal ≠ brain state; medication effects on HRV must be documented |
| BP-07 | EMG Biofeedback (targeted) |
BF-02 (EMG) Specific muscle group amplitude |
Motor rehabilitation · tension · pelvic floor | PW-01, BR-05, BR-10, OUT-01, OUT-02 | Level A | Level 3 demonstrated for targeted applications; mechanism clear at neuromuscular level | Low (for targeted applications) — Signal-muscle causal chain clear; brain-level claims still need caution |
| BP-08 | Respiratory Biofeedback |
BF-05 (Respiratory) Rate, pattern, tidal volume |
Anxiety · asthma · hypertension · vocal | BR-11, OUT-14, OUT-13 (secondary) | Level A–B | Level 3 demonstrated; mechanism via CO₂/pH and baroreflex well-supported | Low–Medium — Direct voluntary control; downstream claims need separate evidence |
| BP-09 | Temperature Biofeedback |
BF-06 (Temperature) Finger temperature increase |
Raynaud's · migraine (peripheral) | BR-13, sympathetic vasomotor, OUT-13 | Level A Raynaud's Level B migraine |
Level 3 demonstrated for Raynaud's; mechanism via sympathetic vasomotor tone clear | Low (for designated applications) — Vasoconstriction mechanism well-characterized for target applications |
| BP-10 | GSR / EDA Biofeedback |
BF-04 (GSR) Skin conductance level |
Arousal reduction | BR-13, sympathetic arousal | Level C | Level 1 demonstrated; clinical Level 3 evidence is weak; primarily research tool | High — EDA cannot distinguish emotional valence; arousal ≠ stress; emotional state inference blocked by R-07 |
These classes extend the Brain Layer uncertainty taxonomy (UA-01–09 and UB-01–13). The same rule applies: no uncertainty class may raise confidence or become positive evidence. Brain Layer inherited classes are referenced by their original IDs.
| Class ID | Class | Cat. | Effect on interpretation | May suspend? | May raise confidence? |
|---|---|---|---|---|---|
| BA-01 | training_mechanism_unconfirmed | BA | Whether feedback training changed the target neural substrate is unconfirmed (Training Level 4 not established) | No — but caps interpretation at Training Level ≤ 3 | Never |
| BA-02 | generalization_unvalidated | BA | Whether within-session signal change transfers to daily-life function is not established (Training Level 3 not demonstrated) | Yes — blocks clinical effectiveness claims | Never |
| BA-03 | operant_conditioning_vs_neural_change | BA | Cannot distinguish whether effect is driven by operant reinforcement, relaxation response, expectancy, or direct neural modulation | No | Never |
| BA-04 | target_signal_nonspecific | BA | The trained signal feature (e.g., alpha power, EDA) is influenced by multiple neural and non-neural processes simultaneously; cannot be attributed to one brain region or mechanism | No | Never |
| BA-05 | within_session_variability_high | BA | Signal feature varies substantially within a session due to state fluctuations, not protocol response; single-session outcome unreliable | No | Never |
| BA-06 | between_session_variability_high | BA | Session-to-session baseline variation exceeds training effect size; learning trend not distinguishable from noise without adequate controls | Yes | Never |
| BA-07 | feedback_demand_characteristics | BA | Person may alter behavior based on perceived expectations of the feedback task, not the specific physiological signal being trained; confounds training interpretation | No | Never |
| CATEGORY BB — BIOFEEDBACK MEASUREMENT UNKNOWNS | |||||
| BB-01 | electrode_impedance_elevated | BB | High scalp-electrode impedance reduces signal fidelity and increases noise floor; results unreliable if impedance >5 kΩ for most EEG systems | Yes | Never |
| BB-02 | emg_crosstalk | BB | EMG signal from adjacent muscles contaminates target muscle recording; muscle specificity claim invalid | Yes | Never |
| BB-03 | reference_electrode_placement_variable | BB | EEG signal is always a differential measurement; reference placement substantially affects power spectra; must be documented | No | Never |
| BB-04 | filter_settings_alter_signal | BB | High-pass, low-pass, and notch filter settings change apparent band power; identical protocol on different devices/settings produces different results | No | Never |
| BB-05 | baseline_drift | BB | Slow DC drift in signal baseline during session causes apparent power changes unrelated to training; particularly affects slow cortical potential protocols | Conditional | Never |
| BB-06 | temperature_ambient_confound | BB | Ambient room temperature independently affects finger temperature and EDA; must be controlled and documented | Conditional | Never |
| BB-07 | sensor_placement_not_standardized | BB | Electrode/sensor placement deviates from protocol specification; cannot compare results across sessions | Yes | Never |
| BB-08 | hair_melanin_fnirs_interference | BB | Hair thickness and melanin concentration substantially affect fNIRS signal quality; must be assessed per individual | Conditional | Never |
| BB-09 | cognitive_demand_confound | BB | The act of watching and responding to the feedback display itself changes the signal (e.g., frontal theta from working memory load); cannot isolate training from feedback-task demands | No | Never |
This extends the Brain Layer v0.1 typed schema. Biofeedback records inherit all Brain Layer fields and add the following. The training_level_demonstrated field is required for all protocol-linked records and defaults to 1 (within-session only).
// Biofeedback extension fields — appended to Brain Layer v0.1 schema { // --- Signal and Protocol --- "signal_id": "enum // BF-01 through BF-08", "protocol_id": "enum // BP-01 through BP-10 | custom | null", "electrode_placement": "string // 10-20 system label(s) or anatomical description", "frequency_band_target": "string | null // e.g. alpha 8-12Hz | SMR 12-15Hz | null for non-EEG", "feedback_metric": "string // the specific scalar feature being fed back, labeled precisely", "feedback_display_type": "enum // auditory_tone | visual_graph | animation | number | reward_token | other", // --- Session Provenance --- "session_number": "int // session number in protocol sequence", "session_duration_min": "int", "electrode_impedance_kohm": "float | null // if EEG; flag BB-01 if >5 kΩ", "device_make_model": "string | null", "software_version": "string | null // flag UB-12 if unknown", "filter_settings": "string | null // highpass, lowpass, notch — flag BB-04 if undocumented", "reference_placement": "string | null // flag BB-03 if undocumented", "ambient_temperature_c": "float | null // flag BB-06 if undocumented for temp/EDA protocols", // --- Training Level Classification --- "training_level_demonstrated": "int // 1=within-session | 2=skill acquisition | 3=generalization | 4=mechanism; DEFAULT 1", "training_level_evidence_source": "string | null // citation or study reference for claimed level", "claimed_level_validated": "bool // false = training level claim not yet validated by cited evidence", // --- Biofeedback Uncertainty (extends brain layer) --- "biofeedback_unknown_class": "enum[] // BA-01 through BA-07 | none", "biofeedback_measurement_class": "enum[] // BB-01 through BB-09 | none", "volume_conduction_applies": "bool // true for all scalp EEG records — BA-04 is active", "generalization_evidence": "enum // absent | anecdotal | weak | moderate | strong", "mechanism_evidence": "enum // absent | proposed | preliminary | supported | confirmed", // --- Inherited scoring gates (override inherited defaults) --- "training_level_claim_valid": "bool // false until training_level_demonstrated ≥ 3 with cited evidence", "person_level_claim_blocked": "bool // inherited from brain layer; always evaluate independently of signal change" }
These rules extend Brain Layer rules R-01 through R-12. All Brain Layer rules remain active and are not repeated here. Rules R-07 (no person-level inference) and R-05 (structure ≠ function) are especially relevant to biofeedback and are inherited unconditionally.
The biofeedback layer cannot be built without the brain layer. All brain region, pathway, and output IDs referenced in this layer depend on the brain layer registry being stable first.
Define what each signal directly measures. Write the signal-to-brain-layer uncertainty map before any protocol records are created. This is the foundation — getting the "what it actually measures" wrong here propagates into all downstream claims.
Define the engineering of the loop before any protocol definitions. Each stage must have its HIR gate logic specified. Feedback display labeling rules (R-21) must be established before any display text is created.
Before any protocol records are created, the training level taxonomy must be locked. Evidence levels must be assigned per protocol and per application — not per modality in general. BCIA, AAPB, and primary literature citations required for any Level B or above claim.
Define each protocol with its signal ID, brain layer links, evidence level, training level evidence, and overclaim risks. No protocol record may assert a Training Level higher than the cited evidence supports.
Formalize BA and BB uncertainty classes as machine-readable objects. Define which classes are active by default for each signal modality (e.g., BA-04 always active for scalp EEG; BB-06 must be documented for temperature and EDA protocols).
Sample records must include: a fully clean EMG motor record, an EEG record with BB-01 triggered, an HRV record with UB-06 triggered, a protocol record where person_level_claim_blocked = true despite valid signal, and a training_level_demonstrated = 3 record with a cited evidence source.
Same requirements as Brain Layer clinical interpretation layer. Biofeedback-specific addition: requires a BCIA-certified or equivalent clinician responsible for protocol selection and outcome interpretation. Protocol output notes may not be presented directly to clients without clinician review. Cannot be a default layer.