Attention, in and Out: Scalp-Level and Intracranial EEG Correlates of Interoception and Exteroception
García-Cordero et al. (2017), from the Ibáñez/Sedeño group (INECO, Buenos Aires), combine three methods that rarely meet — high-density EEG, intracranial depth recording, and behavioural interoceptive learning — on a single question: do interoceptive and exteroceptive attention engage the same attentional machinery, or different mechanisms? The wiki’s earlier García-Cordero citation is the 2016 lesion study on insular-cortex; this is a different paper, and it is the wiki’s first dedicated first-hand HEP study.
The task and the three conditions
The heartbeat-detection task here is a motor-tracking variant (not silent Schandry counting): participants tap a keyboard in time with a signal. Three 2-min conditions:
- Exteroception — track an audio recording of a simulated heartbeat (a constant 60 bpm block and a variable-frequency block).
- Basal interoception — track one’s own heartbeats with no feedback. This is the objective interoceptive accuracy measure (Garfinkel et al. 2015 sense).
- Post-feedback interoception — track one’s own heartbeats again, after a block spent listening to them through a stethoscope. This measures interoceptive learning (sometimes called interoceptive learning or feedback-based improvement).
An accuracy index (0-1) was computed per condition.
What the study found
1. Interoception ≠ exteroception, in the HEP. Relative to exteroception, both interoceptive conditions produced significantly more negative HEP modulations across right, central, and left frontal ROIs in the ~200-500 ms window. This is a within-subject condition contrast (the clean kind for ERP, since it cancels between-person amplitude nuisances), and it says internally- vs externally-directed attention have distinct fast cortical signatures — regardless of whether the person is any good at the task.
2. Interoceptive learning is real behaviourally. Accuracy: exteroception (M=0.73) > post-feedback interoception (M=0.58) > basal interoception (M=0.47), all pairwise significant. One block of stethoscope feedback measurably improved own-heartbeat tracking — but even after learning, tracking an external heartbeat was easier than tracking one’s own, the expected direction if own-heartbeat signals are the weaker channel (see loudness confound).
3. Learning was carried by connectivity, not by HEP amplitude. The HEP did not separate basal from post-feedback interoception. Functional connectivity (weighted symbolic mutual information, wSMI) did: basal interoception = enhanced local, short-range fronto-central connectivity; post-feedback interoception = enhanced long-range, distributed fronto-posterior connectivity. So the study’s reading is that interoceptive learning reconfigures the network — recruiting distant fronto-posterior links tied to memory/multimodal integration — rather than amplifying the primary evoked response. The flat HEP is left as a mild discrepancy with Schandry & Weitkunat (1990), who did find a training-related HEP increase; the authors attribute it to heterogeneous learning aptitude in an unselected sample.
4. Intracranial confirmation, in the frequency domain. In two epileptic patients with depth electrodes in posterior insula, amygdala, somatosensory cortex, and inferior frontal gyrus (pars opercularis) — the canonical interoceptive hubs — basal interoception showed greater broadband high-frequency power (35-110 Hz) and exteroception greater low-frequency power (1-35 Hz), in all four regions. High-frequency broadband activity is associated with internal/effortful, consciousness- and memory-related processing; lower bands with exteroceptive sensory processing. The dissociation is thus not only when (the HEP timing) and where (the frontal scalp), but in what band the two attentional modes operate, recorded directly from human insula.
Where it sits in the wiki
This is the electrophysiological complement to Farb, Segal & Anderson (2013). Farb showed with fMRI that attention moves the interoceptive/exteroceptive boundary within one brain — the same respiratory signal tracked by posterior insula when attended and by somatosensory cortex when a visual task is attended. García-Cordero et al. show the temporal and spectral face of the same claim: attending in vs out changes the frontal HEP within 200-500 ms and flips the dominant frequency band in the insula and its neighbours. Neither says interoceptive and exteroceptive signals are processed by fixed, separate hardware; both say the attentional mode has a distinct, measurable neural signature. See interoception-exteroception-boundary.
It also lands, gently, on the heartbeat-task validity debate from an unusual angle. The HEP dissociation appears irrespective of accuracy — the frontal negativity marks the attempt to attend inward, not success at it. Read one way that is a virtue (a neural index of interoceptive attention that does not require the participant to be a good perceiver, which most are not); read another it underlines that “doing the interoceptive task” and “perceiving your heartbeat” are separable, which is the whole worry. The study does not adjudicate this, and its own accuracy data (basal M=0.47, most participants far from criterion) sit squarely inside the prevalence problem that page documents.
The methodological offering the authors emphasize
The paper’s framing claim is normative: HEP modulation between internal and external attention, being large, within-subject, and reproducible at both scalp and intracranial scales, could serve as a reference parameter against which to read interoceptive HEP abnormalities in psychiatric and neurological conditions — where HEP alterations are reported (Terhaar et al. 2012; Müller et al. 2015; Schulz et al. 2015) but usually against resting baselines or a single interoceptive condition rather than a matched exteroceptive control. See interoceptive-psychopathology.