Article Figures & Data
Figures
- Figure 1.
Schematics for the Interoceptive/Exteroceptive Attention Task (IEAT). In Exteroceptive conditions, participants attended to a circle expanding and contracting; in Interoceptive conditions, participants attended to their respiratory inhalation and exhalation. In Passive conditions, participants simply observed the circle or their breath. In Active conditions, participants pressed buttons to track the circle or breath. In the Matching condition, participants tracked the circle’s movements while synchronizing their respiration cycle with the circle. PasExt = Passive Exteroception; PasInt = Passive Interoception; ActExt = Active Exteroception; ActInt = Active Interoception; ActMatch = Active Matching.
- Figure 2.
Respiration and behavior tracking. A, Respiratory peak and trough detection was successful for each of the experimental task conditions, allowing for calculation of respiratory frequency (Hz) and respiratory volume over time (RVT). B, Respiration was slower during Active Interoception than in other conditions. C, Respiration was deeper (greater RVT) during Active Interoception and Active Matching than in other conditions. D, To assess tracking accuracy, the respiration waveform (blue) was first temporally adjusted (black) to match phase with keypress signals (orange). The average error (red) between key press timing and respiratory peaks/troughs was calculated; the same procedure was applied to a waveform describing circle radius in the exteroceptive tracking conditions. E, No differences in tracking accuracy were observed between Active Interoception of the Breath and Active Exteroception of the Circle, nor with but Active Matching to the Circle; however, breath tracking was superior during the Active Matching condition. The inhibitory effects of RVT on BOLD activity at the first level of analysis are presented in Extended Data Figure 2-1. Ext/Extero = Exteroception; Int/Intero = Interoception.
- Figure 3.
Interactions and simple effects of task. A, Interoceptive Reporting Demand (Active vs Passive) × Attentional Target (Exteroception vs Interoception) implicated the medial and lateral prefrontal cortices (MPFC and LPFC), somatomotor cortices, striatum, and temporal/occipital cortices. B, Median % signal change across the regions revealed highest activation in the Active Exteroception condition and the two Passive conditions relative to Active Interoception and Act Match. C, Simple effects comparisons between Active Exteroception and Interoception conditions largely replicated the interaction effect, suggesting that the Active conditions were driving the interaction. D, Conversely, only area V5/MT distinguished Passive Exteroception from Passive Interoception. Ext = Exteroception; Int = Interoception.
- Figure 4.
Covariates of self-reported interoceptive awareness. A, Orange regions denote positive correlates of MAIA scores within the [ActExt – ActInt] contrast. B, Scatterplot of the relationship between MAIA score and median neural activity during the [ActExt – ActInt] contrast across significant covariate regions. A more detailed scatterplot showing all five experimental conditions is available in Extended Data Figure 4-1, whole-brain maps of MAIA subfactors as covariates are available with TFCE-correction in Extended Data Figure 4-2, and at a more exploratory (p < 0.005, k ≥ 200) threshold in Extended Data Figure 4-3. MAIA scale subfactor intercorrelations and reliabilities are available in Extended Data Table 4-1, factor loadings of these subfactors on a single factor solution are available in Extended Data Table 4-2, and correlations between the subfactors and the Anterior Cingulate Cortex region are available in Extended Data Table 4-3.
- Figure 5.
Endogenous versus exogenous sources of interoceptive control. A, Active Exteroception > Active Matching replicated the deactivation effect observed for Active Interoception. B, Median % signal change across the regions identified via the [Active Exteroception > Active Matching] contrast. C, Active Interoception > Active Matching linked endogenous respiration to greater activation along an insula-operculum pathway relative to the exogenous control condition. D, Median % signal change across the regions identified via the [Active Interoception > Active Matching] contrast. Ext = Exteroception; Int = Interoception.
- Figure 6.
Psychophysiological interaction (PPI) analysis. A, The ACC seed region obtained from the MAIA covariate analysis. B, PPI between ACC regional activity and task conditions [Active Interoception > Active Exteroception], demonstrating enhanced ACC connectivity with a frontoparietal network during Active Interoception relative to Active Exteroception. ACC = Anterior Cingulate Cortex; DLPFC = Dorsolateral Prefrontal Cortex; TPJ = Temporoparietal Junction; ActExt = Active Exteroception; ActInt = Active Interoception.
- Figure 7.
Test-retest reliability of neuroimaging findings. To demonstrate the robustness and replicability of the results, conjunction analysis for the four main contrasts reported in this paper was conducted. As the TFCE algorithm does not currently perform conjunction analysis, and halving the sample size reduces experimental power, the baseline and postintervention sessions were each analyzed separately at p < 0.0011/2 = p < 0.0316, so that the resulting overlap would yield a conjunction p < 0.001; a cluster size of k ≥ 500 was also applied to each estimate as this cluster size resulted in a FWE corrected p value < 0.05 for p < 0.001. A, the contrast of Active Exteroception – Active Interoception reproduced the pattern of somatomotor and prefrontal activation. B, The MAIA covariate of [Active Exteroception – Active Interoception] reproduced the anterior cingulate region of interest. C, The contrast of Active Interoception – Active Matching reproduced greater insula/opercular activation for endogenously-paced interoception (Active Interoception) than exogenously-paced interoception (Active Matching). D, The PPI analysis reproduced increased connectivity between the anterior cingulate and the dorsal attention network (DAN). DMPFC = dorsomedial prefrontal cortex; DLPFC = dorsolateral prefrontal cortex; ACC = anterior cingulate.
Tables
95% CI Condition Mean SE df Lower Upper Active Interoception 0.188 0.0116 24.6 0.164 0.211 Passive Interoception 0.197 0.0116 24.6 0.173 0.221 Active Matching 0.209 0.0116 24.6 0.185 0.233 Active Exteroception 0.212 0.0116 24.6 0.188 0.235 Passive Exteroception 0.212 0.0116 24.6 0.189 0.236 Data are displayed in Hz, SE = standard error of the estimate, df = degrees of freedom, CI = confidence interval.
Contrast Estimate SE df t value p value ActExt – ActInt 0.024 0.00507 854 4.745 <0.0001* ActExt – ActMatch 0.002 0.00507 854 0.449 0.9916 ActExt – PasExt −0.001 0.00507 854 −0.187 0.9997 ActExt – PasInt 0.015 0.00507 854 2.921 0.0294* ActInt – ActMatch −0.022 0.00507 854 −4.296 0.0002* ActInt – PasExt −0.025 0.00507 854 −4.932 <0.0001* ActInt – PasInt −0.009 0.00507 854 −1.824 0.3601 ActMatch – PasExt −0.003 0.00507 854 −0.636 0.9692 ActMatch – PasInt 0.013 0.00507 854 2.472 0.098 PasExt – PasInt 0.016 0.00507 854 3.108 0.0166* ActExt = Active Exteroception; ActInt = Active Interoception; ActMatch = Active Matching; PasExt = Passive Exteroception; PasInt = Passive Interoception. * indicates p values < .05.
95% CI Condition Mean SE df Lower Upper Active Exteroception 2.43 0.0407 360 2.35 2.51 Passive Exteroception 2.51 0.0407 360 2.43 2.59 Passive Interoception 2.72 0.0407 360 2.64 2.80 Active Matching 3.15 0.0407 360 3.07 3.23 Active Interoception 3.42 0.0407 360 3.34 3.50 Data are reported in respiratory volume/time, in terms of respiratory belt stretch (arbitrary units) integrated over time.
Contrast Estimate SE df t value p value ActExt – ActInt −0.99 0.0573 854 −17.268 <0.0001* ActExt – ActMatch −0.713 0.0573 854 −12.44 <0.0001* ActExt – PasExt −0.078 0.0573 854 −1.361 0.6527 ActExt – PasInt −0.289 0.0573 854 −5.046 <0.0001* ActInt – ActMatch 0.277 0.0573 854 4.828 <0.0001* ActInt – PasExt 0.912 0.0573 854 15.907 <0.0001* ActInt – PasInt 0.701 0.0573 854 12.222 <0.0001* ActMatch – PasExt 0.635 0.0573 854 11.079 <0.0001* ActMatch – PasInt 0.424 0.0573 854 7.394 <0.0001* PasExt – PasInt −0.211 0.0573 854 −3.685 0.0023* Data are reported in respiratory volume/time, in terms of respiratory belt stretch (arbitrary units) integrated over time. ActExt = Active Exteroception; ActInt = Active Interoception; ActMatch = Active Matching; PasExt = Passive Exteroception; PasInt = Passive Interoception. * indicates p values < .05.
95% CI Target Condition Mean SE df Lower Upper Breath Matching 179 28.5 28 120 237 Circle Matching 231 28.5 28 172 289 Circle Exteroception 239 28.5 28.1 180 297 Breath Interoception 262 28.6 28.3 204 321 Target indicates the target for which button tracking accuracy is calculated; for Active Exteroception, this is always the circle, and for Active Interoception, this is always the breath, but for the Active Matching condition, both targets are required to align with keypresses, so both accuracies can be computed. Data reported are in milliseconds lag relative to the target inflection points.
Target Condition Target Condition Estimate SE df t value p value Circle Exteroception vs Circle Match 7.93 17.1 673 0.463 0.967 Breath Interoception vs Circle Exteroception 23.68 17.3 673 1.372 0.5176 Breath Interoception vs Circle Match 31.61 17.2 673 1.834 0.2583 Breath Match vs Breath Interoception −83.79 17.2 673 −4.861 <0.0001* Breath Match vs Circle Exteroception −60.11 17.1 673 −3.508 0.0027* Breath Match vs Circle Match −52.17 17.1 673 −3.05 0.0127* Data reported are in milliseconds lag relative to the target inflection points. * indicates p values < .05.
Cluster Peak p Peak Peak Peak p Coordinates
(MNI)Description Side size (k) (FWE) TFCE Z (raw) x y z Interaction [ActExt > ActInt] > [PasExt > PasInt] Somatomotor, middle temporal,
superior frontal, temperoparietal junction- 34501 0.001 2389.64 3.54 <0.001 −50 −32 56 Occipital pole L 231 0.037 1181.47 3.24 0.001 −4 −94 30 Angular gyrus R 226 0.043 1136.62 3.16 0.001 58 −54 20 Insula L 169 0.043 1134.44 3.09 0.001 −22 8 −2 Inferior frontal R 70 0.046 1113.62 3.09 0.001 50 24 18 ActExt > ActInt Somatomotor, middle temporal, superior frontal,
temperoparietal junction- 57088 0.001 2540.6 3.54 <0.001 −50 −32 56 Occipital pole L 318 0.032 1249.39 2.95 0.002 −22 −94 12 Occipital pole R 55 0.043 1152.73 3.16 0.001 16 −96 16 PasExt > PasInt Area V5/MT R 62 0.033 1057.65 3.09 0.001 46 −64 4 ActExt = Active Exteroception; ActInt = Active Interoception; ActMatch = Active Matching; PasExt = Passive Exteroception; PasInt = Passive Interoception.
Cluster Peak p Peak Peak Peak p Coordinates
(MNI)Description Side size (k) (FWE) TFCE Z (raw) x y z Anterior cingulate L 1205 0.025 1677.36 2.85 0.002 −4 12 36 Postcentral (Somatosensory) L 921 0.03 1594.48 2.79 0.003 −60 −10 32 Precentral (motor) R 122 0.036 1502.99 2.88 0.002 28 −14 78 Dorsolateral PFC/Broca’s L 489 0.037 1480.51 2.99 0.001 −34 32 56 Lateral occipital/Wernicke’s L 1047 0.038 1467.38 2.95 0.002 −52 −78 −20 Lateral occipital R 160 0.039 1464.02 2.88 0.002 54 −76 −12 Dorsomedial PFC R 61 0.042 1424.77 3.09 0.001 20 46 54 Insula L 87 0.045 1386.84 2.71 0.003 −42 −8 −8 Cerebellum R 77 0.046 1381.72 2.91 0.002 54 −74 −36 Cerebellum R 74 0.047 1367.92 2.88 0.002 38 −44 −40 Dorsolateral PFC L 119 0.047 1367.11 2.79 0.003 −30 54 30 Lateral occipital (Superior) R 32 0.047 1362.52 2.71 0.003 10 −82 44 Cerebellum R 35 0.048 1358.1 2.95 0.002 24 −46 −50 Frontal pole R 26 0.048 1355.7 2.64 0.004 26 70 10 Orbitofrontal L 13 0.048 1351.04 2.73 0.003 −56 24 −10 ActExt = Active Exteroception; ActInt = Active Interoception; ActMatch = Active Matching; PasExt = Passive Exteroception; PasInt = Passive Interoception.
- Table 9
MAIA subscale correlates with MAIA covariate ROI: means, SDs, and correlations with confidence intervals
Variable M SD Correlation 95% CI Total score 15.13 5.35 −0.32* [−0.56, −0.03] Nonjudging 26.73 7.59 −0.16 [−0.43, 0.15] Noticing 2.82 1.07 −0.14 [−0.42, 0.17] Attention regulation 2.06 1.07 −0.27 [−0.52, 0.03] Emotion awareness 3.22 0.96 −0.32* [−0.56, −0.02] Self-regulation 2.37 1.10 −0.28 [−0.53, 0.02] Body listening 1.77 1.20 −0.28 [−0.53, 0.02] Trusting 2.88 1.25 −0.27 [−0.52, 0.03] M and SD are used to represent mean and standard deviation, respectively. Values in square brackets indicate the 95% confidence intervals. * indicates correlations with p values < .05.
Cluster Peak p Peak Peak Peak p Coordinates
(MNI)Description Side size (k) (FWE) TFCE Z (raw) x y Z ActExt > Act Match Cerebral cortex - 134,079 0 6116.73 3.54 <0.001 −58 −24 16 ActInt > Act Match Insula, operculum, ventral striatum, lingual gyrus R 54,630 0 3898.16 3.54 <0.001 60 10 −2 Dorsolateral PFC L 706 0.011 1435.25 3.54 <0.001 −32 54 26 Middle temporal gyrus L 10 0.049 952.43 3.54 <0.001 −52 −42 −6 ActExt = Active Exteroception; ActInt = Active Interoception; ActMatch = Active Matching.
Cluster Peak p Peak Peak Peak p Coordinates
(MNI)Description Side size (k) (FWE) TFCE Z (raw) x Side Size (k) Dorsal attention network: frontal R 23783 <0.001 3269.41 3.54 0 44 24 40 Dorsal attention network: parietal R 18671 0.001 2368.12 3.54 0 58 −40 52 The PPI was defined by the interaction of the [ActInt – ActExt] contrast * ACC activity.
Extended Data Figure 2-1
Effects of the respiratory volume/time (RVT) covariate on BOLD activity. Contrasts maps were generated from the five-condition model corrected using threshold-free cluster estimation (TFCE) to the familywise p < 0.05 level. Areas in blue demonstrate reduced BOLD activity with greater levels of RVT. Download Figure 2-1, TIF file.
Extended Data Figure 4-1
MAIA scale and anterior cingulate (ACC) activation with all five task conditions. The relationship between interoceptive sensibility/awareness scores (MAIA) and anterior cingulate activity across all five task conditions. Signal was extracted from each experimental condition and plotted above, demonstrating the same reduced separation between conditions, including within the passive conditions. ActExt = Active Exteroception, ActInt = Active Interoception, ActMatch = Active Matching, PasExt = Passive Exteroception, PasInt = Passive Interoception. Download Figure 4-1, TIF file.
Extended Data Figure 4-2
MAIA covariates of the [Active Interoception – Active Exteroception] contrast, corrected using threshold-free cluster estimation (TFCE) to the familywise p < 0.05 level. The analysis was run on each MAIA subscale in place of the total score as a covariate in a model that also controlled for study design (Group × Time), respiration frequency, and RVT. Only covariates with significant voxels are displayed; only Emotional Awareness and Self-Regulation showed evidence of covariation. Emotional Awareness also showed a TFCE-corrected cluster of activity in the left planum temporal/superior temporal gyrus. Self-Regulation also showed a TFCE-corrected cluster of activity in the right lateral cerebellum at the border of Crus1 and Crus2. Download Figure 4-2, TIF file.
Extended Data Figure 4-3
MAIA covariates of the [Active Interoception – Active Exteroception] contrast, corrected using the joint thresholds of p < .005 and cluster size k ≥ 200 voxels. The analysis was run on each each MAIA subscale in place of the Total score as a covariate in a model that also controlled for study design (Group x Time), respiration frequency, and RVT. Only covariates with significant voxels are displayed. As illustrated above, many components shared substantial overlap with the pattern observed in the MAIA total score. Emotion Awareness, Self-Regulation, Body Listening, and to a lesser extent Attention Regulation and Noticing all showed similar patterns of covariation, a pattern that aligns with the ACC ROI correlations described in Table 4-3. These whole brain analyses do suggest that Emotion Awareness, Self-Regulation, and Body Listening also implicate the left lateralized language regions, suggesting a particular role of the “Mind-Body Integration” conceptual domain in supporting the covariation effect. Noticing only implicated the temporal and somatosensory association aspects of the total score covariate, and conversely, Attention Regulation implicated the ACC without many of the left-temporal language regions. There was little overlap with Trusting, which only implicated the left frontal pole and right cerebellum (Crus 1), Not Worrying, which only implicated the right superior occipital cortex, and no covariation with Not Distracting even at this exploratory threshold. Download Figure 4-3, TIF file.
Extended Data Table 4-1
MAIA Scale subfactor intercorrelations and reliabilities. Download Table 4-1, DOCX file.
Extended Data Table 4-2
Factor loadings for a single factor solution of the eight MAIA subfactors. Download Table 4-2, DOCX file.
Extended Data Table 4-3
Univariate Correlations of Anterior Cingulate ROI with MAIA total score and subscales. Download Table 4-3, DOCX file.
In this issue
