Social Antecedents to the Development of Interoception: Attachment Related Processes Are Associated With Interoception

Study at a glance

• Authors: Kristina Oldroyd, Monisha Pasupathi, Cecilia Wainryb
• Year: 2019
• Journal: Frontiers in Psychology
• DOI: 10.3389/fpsyg.2019.00712
• N: Study 1: 135 young adults (68 male); Study 2: 108 youth (53 male, ages 8-17) and their mothers
• Design: Two cross-sectional secondary analyses of existing datasets — Study 1: self-report questionnaire correlations (MAIA x ECR-S); Study 2: psychophysiology (skin conductance) + self-report congruence, regressed on maternal parenting-style questionnaire

Key findings

• Attachment anxiety correlated positively with MAIA Noticing (r = 0.18) and Emotional awareness (r = 0.18), and strongly negatively with Not-worrying (r = -0.43) — a hypervigilant, worried relationship to bodily cues.
• Attachment avoidance correlated negatively with MAIA Attention regulation (r = -0.20) and Trusting (r = -0.26) — less attention to bodily cues and less trust in them as reliable signals.
• Maternal rejection of children’s negative emotion predicted lower self-report/physiology congruence in youth (beta = -0.21, r = -0.26), i.e. the more a mother rejects her child’s negative emotion, the less the child’s reported distress tracks their sympathetic arousal.
• Maternal acceptance of negative emotion did not predict higher congruence, contrary to hypothesis H2b — the authors conclude the acceptance subscale may not index sensitive caregiving (it was uncorrelated with self-reported maternal warmth, whereas rejection was).
• Frames interoception as having social antecedents: proposed mechanisms are attachment-driven HPA-axis dysregulation (altering interoceptive signal strength and processing), attachment effects on the growth of the interoceptive network (AIC/ACC/OFC), and caregiver mirroring in the construction of the bodily self.

Limitations

• Both studies are secondary analyses of datasets collected for unrelated questions (embodied narration and caloric consumption; narration and anger regulation) — no measure was chosen for this paper’s hypotheses.
• Entirely cross-sectional and correlational despite a developmental thesis: no child was followed over time, and Study 1’s ‘development’ claim rests on adult self-reports of current attachment style.
• Study 1 measures both constructs by self-report in a college sample, so shared method variance is uncontrolled; the Not-worrying/anxiety correlation (r = -0.43) plausibly reflects trait negative affectivity common to both scales.
• Most a priori hypotheses failed. Of H1a’s six predicted avoidance effects only Trusting held; H1b predicted avoidant individuals would score higher on Attention regulation and the observed effect ran the opposite way; H1d’s predictions were largely null. The paper reports these as support without flagging the reversal.
• Effects are small and the Study 2 model is weak (R-squared ~0.05-0.08, F = 2.92, p = 0.05); Bonferroni adjustment is claimed but the reported alpha thresholds appear unadjusted.
• The congruence score is an unvalidated interoception proxy — the authors call it ‘imperfect’ and list validating it as future work. It rests on a single citation (Kleckner et al. 2017) and is computed from only five within-person data points.
• Numerous internal reporting inconsistencies (see ‘Data-reporting problems’ below) — the abstract, text, and tables disagree on sample size, effect sizes, and even the direction of the headline Study 2 result.

Connections

• Methods: Multidimensional Assessment of Interoceptive Awareness (MAIA), physiology congruence, Heartbeat detection task
• Concepts: Interoception, Attachment style, Social origins of interoception, HPA axis, Embodied selfhood, Interoceptive taxonomy (Farb et al. Box 1), Interoceptive sensitivity, Insular cortex, Anterior cingulate cortex (ACC)
• Debates: Are the origins of interoception social or biological?
• Related studies: Craig (2002), Seth (2013), Farb et al. (2015)

The first paper in this wiki to ask where interoception comes from. Every other source treats interoceptive ability as a given — an anatomical endowment (Craig), a precision parameter (Seth), or a trainable skill (Farb et al.). Oldroyd et al. ask what builds it, and answer: early caregiving. See Social origins of interoception.

Open access (CC BY), Frontiers in Psychology 10:712.

The argument

Interoceptive brain regions (AIC, ACC, OFC) show protracted post-natal development, which “leaves a substantial window of opportunity for environmental input.” Since attachment is the dominant environmental input in that window, Attachment style should shape interoception. Three proposed mechanisms:

1. Signal strength and processing via the HPA axis. Attachment styles differ in HPA reactivity to stress. Because stress (descending) and interoception (ascending) run through the same brain-body pathways, HPA dysregulation could alter both the strength of the interoceptive signal (via stroke volume — Schandry et al. 1993 link higher stroke volume to better heartbeat detection) and its processing (cortisol lowers the threshold for interoceptive signal processing; Schulz et al. 2013).
2. Neural architecture. Anxious/avoidant children show lower insular volume and surface area (Kühn & Gallinat 2013; Lim et al. 2014), and avoidant adults show decreased insular activation to stimuli (DeWall et al. 2011). See Insular cortex, Anterior cingulate cortex (ACC).
3. The bodily self. Infants cannot act to test the causes of their own interoceptive states, so they depend on caregivers’ responses to learn what their sensations mean — the infant learns “hunger” only when a caregiver’s nipple resolves the discomfort (following Fotopoulou & Tsakiris 2017; Stern 1985). See Embodied selfhood.

Study 1 — attachment style x MAIA (N = 135)

Self-reported attachment (ECR-S: anxiety, avoidance) against the eight Multidimensional Assessment of Interoceptive Awareness (MAIA) subscales.

	Noticing	Not-distracting	Not-worrying	Attention reg.	Emotional awareness	Self reg.	Body listening	Trust
Anxious	0.18*	-0.05	-0.43**	-0.05	0.18*	-0.12	0.05	-0.09
Avoidant	-0.14	-0.10	-0.03	-0.20*	-0.04	-0.13	-0.12	-0.26**

The interpretation: people’s responsivity to bodily cues mirrors their responsivity to interpersonal cues. Anxiously attached individuals, hypervigilant to relational threat, are correspondingly hypervigilant to bodily threat — they notice more and worry more. Avoidantly attached individuals, who preemptively disengage attention from distressing social cues, likewise divert attention from bodily cues and do not trust them (“a person who has not developed trust in a loving caretaker… may not expect their body to give them reliable and important signals”).

Note the pattern is not “avoidant = uniformly worse interoception.” Only 2 of 8 subscales reached significance for avoidance, and 3 of 8 for anxiety. The anxious profile is elevated on some dimensions.

Study 2 — parenting style x self-report/physiology congruence (N = 108 dyads)

A tighter test, using an objective channel. Youth recalled an anger memory across five epochs (vanilla baseline, talking baseline, exposure, regulation, re-exposure) while skin conductance level was recorded; after each epoch they rated negative emotion. The congruence score is the within-person correlation between the two five-point series (observed range: r = -0.84 to 1.0). Mothers separately completed the Emotion Related Parenting Scale-Short.

Result: maternal rejection of negative emotion predicted lower congruence (beta = -0.21, p < 0.05; zero-order r = -0.26). Maternal acceptance did not predict higher congruence (beta = 0.05, n.s.).

The authors read the rejection finding as the developmental analogue of the classic avoidant psychophysiological profile — minimized self-reported distress alongside elevated physiological distress (Dozier & Kobak 1992; Roisman et al. 2007). The clearest statement of the thesis is the vignette: a parent who says “Ouch! That must have hurt” promotes interoception; a parent who says “You’re fine! That didn’t hurt! Get back up!” teaches the child to distrust their own body. Following Stern (1985) and Fonagy (2001) — for a child to know their own mind they must see it reflected in a sensitive caregiver — the authors contend that for a child to know their own body, they must see it reflected too.

The null for acceptance is handled honestly: the authors report that the acceptance subscale was uncorrelated with maternal warmth while the rejection subscale was (r = -0.18, p = 0.04), and conclude the acceptance scale is probably a poor index of sensitive caregiving rather than that acceptance doesn’t matter.

Data-reporting problems

Flagged because they bear on how much weight the numbers can carry. These are internal inconsistencies within the published paper, not disagreements with other wiki sources:

• Sample size is inconsistent. The abstract says n = 132 and reports r’s(130); the Methods say 135 participants; the Results report r(133) (implying N = 135). Table 2 says N = 135.
• The abstract states the Study 2 headline result backwards. It reads “The congruence score was positively associated with parental rejection of negative emotion” and then, two sentences later, reports r(108) = -0.24 and describes the relation as negative. The body of the paper is consistently negative; the abstract’s first sentence is an error.
• The Study 2 effect size differs in all three places it appears: r = -0.24 (abstract), r = -0.26 (Table 4), beta = -0.21 (Table 5).
• The manipulation-check ANOVAs are impossible as printed: “MAIA: F(3,129) = 11.96, p = 0.41” and “ECRS: F(3,131) = 42.20, p = 0.94.” An F of 11.96 on (3,129) df has p < 0.001, not 0.41. Either the F or the p is wrong in both.
• The Study 2 regression df don’t match the model. Four predictors are entered but the test is reported as F(1,107) = 2.92; it should be F(4,103). R-squared is given as 8.4% (adjusted 4.7%) in the text but 0.05 in Table 5.
• The regression equation is malformed: “Congruence score = 0.33*(-0.03) (Mother’s score on parental rejection of emotion scale)” is not a well-formed equation.
• MAIA scoring is internally inconsistent. The text says items are coded 0-5 but subscale scores “can vary in the 1-5 range”; Table 2 gives Not-distracting and Trust a maximum of 3 (vs. 5 for the other six subscales) and a MAIA total of 20-38, which is not reconstructible from averaged 1-5 subscales.
• Table 1 disagrees with itself: the Not-worrying x Attention-regulation cell is 0.32* above the diagonal and 0.32** below it.

None of this touches the direction of the two headline findings, which are consistent across text and tables (bar the abstract slip). But the paper should be cited for its framing and its direction of effect, not for precise effect sizes.

Why this matters to the wiki

• It supplies the developmental layer the wiki was missing. The Are the origins of interoception social or biological? debate opens here: Fotopoulou & Tsakiris’s claim (quoted approvingly) that the origins of core feelings like hunger and warmth are “social not biological” sits in real tension with Craig’s phylogenetic, homeostatic-afferent framing of the same feelings.
• It is the wiki’s first source using the Multidimensional Assessment of Interoceptive Awareness (MAIA), which Farb et al.’s taxonomy classifies under sensibility — worth remembering when reading these correlations, since none of them touch interoceptive accuracy.
• It gives an aetiology for the avoidant profile that resembles alexithymia and the clinical presentations in Interoceptive training in clinical practice, suggesting where such profiles come from rather than only how to treat them.