Lyons et al. (2021) — Bodily Maps of Emotion in Major Depressive Disorder
The first application of emBODY to depression, and the wiki’s first source in which the bodily-maps paradigm is used clinically rather than to argue about universality. Three groups of 30 — never-depressed controls (HC), current MDD off antidepressants (MDDnm), current MDD on antidepressants (MDDm) — coloured where they felt activation and deactivation while viewing IAPS pictures cued to anger, disgust, happiness, sadness, fear, and neutral.
Two aims: conceptually replicate Nummenmaa et al. (2014) with a new induction method, and ask whether depression changes the maps. An exploratory third aim — the effect of antidepressant medication — produced the most interesting result.
The negative result is the good one
The authors predicted, extrapolating from Torregrossa et al. (2018), that depressed participants’ BSMs would be indistinguishable between emotions, as they are in schizophrenia. They are not. Classification accuracy in both patient groups matched controls (34–36%, and 43–45% once anger is dropped), and the confusion matrices look like the control matrix. Better still, the classifier trained on controls transferred to both patient groups above chance.
That is a cleaner finding than the abstract’s framing suggests, and it points somewhere specific: depression does not relocate emotion in the body. Whatever is wrong with emotion in depression, it is not that sadness has moved, or that the felt body has become undifferentiated mush. Schizophrenia does that; depression does not. The authors’ own conclusion — “emotion experience seems to be differentially affected by the two mental disorders” — is the durable claim here, and it is a dissociation between two disorders on one instrument, which is worth more than either group difference.
The arithmetic problem — read this before citing any number from this paper
Table 2’s three ratio columns are labelled with the groups reversed. This is not an inference; it is checkable against the paper’s own pixel counts, and all 18 cells match to two decimals under the transposed reading:
| column as printed | what the numbers actually are |
|---|---|
| ”Activation of MDDnm relative to HC in %“ | MDDm / HC |
| ”Activation of MDDm relative to HC in %“ | MDDnm / HC |
| ”Activation of MDDm relative to MDDnm in %“ | MDDnm / MDDm |
Worked example (Fear): printed “MDDnm relative to HC = 23.52%”, but 888 / 61,144 = 1.45%, which is the number printed one column to the right; and 14,384 / 61,144 = 23.52%, i.e. MDDm/HC. Same pattern in all six rows of all three columns.
The discussion inherits the error. It reports “MDDnm painted only 63% as many pixels as those of HC overall, i.e., colored around half as many pixels as HCs” — a sentence that is internally inconsistent (63% is not “around half”) because the two figures describe different groups. Recomputing from the pixel counts:
- MDDnm painted 71.2% of HC’s pixels overall.
- MDDm painted 49.9% — that is the group that “colored around half as many pixels as HCs”.
So the “around half” observation is real but belongs to the medicated group, which fits the medication story, not the depression story.
What survives, and what does not
Under the corrected reading, MDDnm relative to HC by emotion is: neutral 285%, disgust 137%, sadness 96%, anger 79%, happiness 73%, fear 1.45%. Unmedicated depressed participants painted more than controls for neutral and disgust, and essentially the same for sadness.
Excluding fear, MDDnm painted 104% of HC — very slightly more. The entire “less overall activation in depression” claim rests on the fear cell: 888 pixels against 61,144, a 69-fold gap where no other cell in the table exceeds a 3-fold difference. That cell is a post-FDR survivor count at a threshold (2.56) chosen per-group, in the emotion the authors elsewhere report they failed to induce (“the higher misclassification of fear as a neutral emotional state might be caused by a failed induction of fear through mildly arousing pictures”). An 888-pixel result is as consistent with a thresholding artifact as with blunted fear. The paper does not remark on the magnitude anywhere.
The abstract’s “MDDnm showed less overall activation than HCs, especially in sadness and fear” therefore needs splitting. The sadness part does not come from the pixel table — sadness is at 96% — but from the FDR-thresholded difference maps (Fig. 1d), which localise reduced stomach and chest activation. A localised difference alongside an equal total is perfectly coherent, and is the better-supported half of the claim. The “less overall activation” part is the half that does not survive the corrected arithmetic.
Recorded as an internal defect of the source, not as a wiki contradiction: no page here asserted anything about Lyons before this ingest. But it means this study should not be cited for “depression blunts bodily emotion” without the caveat, and the is-more-interoceptive-awareness-better debate leans on it only lightly for that reason.
The medication finding, which the arithmetic does not touch
The MDDm result is a within-group proportion and so is immune to the transposition: 63.34% of the medicated group’s painted area was deactivation, against 38.76% (MDDnm) and 24% (HC), and MDDm was the only group in which deactivation survived FDR at all. Difference maps show less activation everywhere except heart, stomach and head.
The authors connect this to reports of emotional numbing on antidepressants (Read et al. 2014: 60% of 1829 New Zealanders; Price et al. 2009; Kajanoja et al. 2018) — see antidepressant-emotional-blunting. It is a genuinely novel observation: a topographical correlate of a side effect otherwise known only from questionnaires, in people who can still tell their emotions apart.
Two cautions the wiki should keep attached. First, the authors’ own: the medicated group might have been more severely depressed before treatment, so the groups may not be comparable despite equal current BDI-II. Second, one they do not raise: because emBODY folds activation and deactivation into a single map, “more deactivation” and “less painting overall” are not separable here — and MDDm did paint least overall (49.9% of HC). A group that paints less and paints it blue may be reporting numbness, or may be reporting less of everything.
Where it sits in this wiki
On the paradigm. This is the first independent group in the wiki to run emBODY and replicate the core Nummenmaa result — with a different induction (pictures), a different language (German), and a clinical sample. That matters for Nummenmaa’s programme, since replication by a lab with no stake in it is worth more than another in-house extension. It also quietly strengthens the embody page’s note about the word-cue problem: these maps were cued by pictures, not words, and still came out emotion-specific.
On the deactivation discrepancy. In HC, “almost no deactivation reached significance” — unlike Nummenmaa’s original maps, where deactivation is a substantial part of the signature (and where, per volynets-2020-cultural-universality, sadness and depression are carried primarily by deactivation). The authors note the discrepancy and let it stand. If the deactivation half of a BSM depends this much on induction method, then the “fingerprint” is partly an artifact of how it is elicited — a real problem for the constitutive claim that BSMs determine emotional qualia, and now logged on bodily-sensation-maps.
On the ageing tension. Volynets et al. (2020) — same instrument, same senior-author lineage — propose that reduced felt bodily emotion in ageing is benign or even beneficial, making emotion easier to regulate (age-related-interoceptive-decline). Lyons et al. find reduced felt bodily emotion in depression and conclude that increasing body awareness “might be an important target in the treatment of patients with major depression.” Same measure, opposite valuation. That tension now has a home at is-more-interoceptive-awareness-better, where the resolution is unlikely to be that one paper is wrong.
On the clinical recommendation. The conclusion — support body-awareness approaches (MBCT, emotion-focused therapy, exercise) in depression — runs directly into Farb et al.’s (2015) explicit caution that enhancing interoceptive awareness without a regulatory framework can be maladaptive in severely depressed and suicidal patients specifically. Lyons et al. excluded currently suicidal patients, so their sample is precisely the one Farb et al.’s warning does not cover, and their data cannot speak to it. See interoceptive-training-clinical.