Method at a glance
- Modality: gastric
- Measures: The stomach’s slow myoelectrical rhythm (~0.033–0.066 Hz, i.e. 2–4 cycles per minute), recorded non-invasively from cutaneous electrodes on the abdomen. Yields a peak gastric frequency and a continuous phase signal — a physiological measure of the organ, not a perceptual measure of the person.
- Procedure: Four pairs of cutaneous electrodes on the abdomen plus a hip-bone ground, supine, after a fast (~3 h food, 0.5 h drink). Amplify (gain 5000, 0.005–1 Hz band), downsample, then select the channel with the largest spectral power peak inside the normal gastric range using Welch’s method; band-pass around that peak (±0.015 Hz) and Hilbert-transform to extract instantaneous phase. MRI-compatible amplifiers permit concurrent recording with fMRI.
- Also called: EGG, cutaneous electrogastrography, gastric myoelectrical recording
Strengths
- Non-invasive and continuous. Gives an uninterrupted phase signal at rest, requiring nothing of the participant — no report, no judgment, no task. It measures a body, not an opinion about a body.
- Highly reliable at the signal level. Peak gastric frequency retested at r(14) = .74 across recordings taken months apart in different settings (Levakov et al. (2023)) — better stability than most interoceptive measures in this wiki.
- A slow, stable rhythm. ~20 s per cycle with an individually characteristic frequency, which makes phase-based analysis against fMRI tractable in a way cardiac timing (~1 Hz, far above the BOLD band) is not.
- Normative data and a published quality-assessment pipeline exist (Wolpert, Rebollo & Tallon-Baudry 2020), which is more methodological infrastructure than several better-known interoceptive tasks have.
Limitations
- High and non-random data loss. ~24% of runs are unusable for poor signal quality (29/122 in Levakov et al. (2023); 21/117 in Wolpert et al. 2020). Whoever is excluded is excluded for a physiological reason, so the surviving sample is not a random subsample.
- BMI-restricted by construction. Signal quality degrades with abdominal adiposity, so studies routinely restrict to BMI 18–25. This silently removes the population for which gastric interoception is most clinically interesting.
- Confounded with respiration. Breathing modulates the gastric signal and shares part of its frequency neighbourhood (Verhagen et al. 1999; Yin & Chen 2013), and also moves the head — which is how Gastric network estimates acquire motion artefact that is correlated with the signal of interest rather than independent of it.
- Electrode placement is uncalibrated. Standard practice places electrodes by external landmark, not by imaging the stomach; high-density and imaging-guided variants exist (Gharibans et al. 2019) but are not the norm.
- It measures the organ, not the perception. EGG belongs to the physiology column of the Interoceptive taxonomy (Farb et al. Box 1). It says nothing about whether the participant can detect, discriminate, or report anything gastric — a distinction easy to lose when EGG-derived quantities are called ‘gastric interoception’.
- Hunger state modulates the signal and is rarely standardized beyond a fasting instruction.
Connections
The gastric counterpart to the ECG, and the reason the Gastric network is knowable at all. Four electrode pairs on the abdomen pick up the smooth-muscle depolarizations driven by the interstitial cells of Cajal — the stomach’s pacemaker — at 2–4 cycles per minute.
The clinical use is decades old (gastroparesis, functional dyspepsia; Parkman et al. 2003). The interoception use is recent and different in kind: EGG is recorded concurrently with fMRI, not to diagnose the stomach, but to use the stomach’s rhythm as a reference signal against which brain activity can be phase-aligned.
Where it sits among this wiki’s instruments
Almost every method page here measures a judgment — how many beats did you count, which breath was harder, where do you feel anger. EGG measures none. It is one of a small set of pages (Event-related potentials partly, physiological recording generally) where the participant supplies a body and no opinion.
That has a consequence the field has not fully absorbed. The reliability of interoceptive tasks is contested throughout this wiki — see Is the heartbeat-counting task a valid measure of cardiac perception?. The reliability of EGG is not: r = .74 across months and settings. What fails to retest is not the visceral measurement but the coupling quantity computed from it and the BOLD signal (PLV-delta, r ≈ 0). Whatever is unstable in brain–body coupling research, it is not the stomach.
The pipeline, and where choices bite
- Channel selection. Four channels are recorded; one is chosen — the largest spectral peak inside 0.033–0.066 Hz, unless a quality assessment favours the second-largest. A researcher degree of freedom, usually reported briefly.
- Quality grading. Wolpert et al. (2020) supply three criteria (clear peak in range; peak consistent across channels; peak consistent across runs) and three grades; grade 3 is discarded. This is the source of the ~24% attrition.
- Band-pass and Hilbert. Filter ±0.015 Hz around the individual peak, then take the analytic signal’s angle to get instantaneous phase. Everything downstream is phase; amplitude is discarded.
- Downsampling to the BOLD rate before computing PLV against each voxel.
Step 3 is why this method pairs with PLV specifically: what EGG delivers cleanly is when in the gastric cycle you are, not how strong the contraction is.
What it would take to make it a measure of a person
Presently EGG supports group-level claims well and individual-difference claims badly, and the failure is downstream of the recording. Improvements the literature points at: imaging-guided electrode placement, high-density arrays (Gharibans et al. 2019), automated artefact rejection (Gharibans et al. 2018), and — the one that would matter most for this wiki — a perceptual gastric task to sit beside the physiological one, so that the gastric channel could enter Is interoception domain-general or organ-specific? as a third axis with matched psychophysics rather than as an absence.