KurisutinaCompanion to the main-paper summary tasserie2022_thalamic_dbs.md. This note records the supplementary reading assigned separately to the connectomics worker; the consciousness worker read and summarized the complete main article. It is not a separate primary paper.
Source: Supplementary Materials for “Deep brain stimulation of the thalamus restores signatures of consciousness in a nonhuman primate model,” Science Advances 8, eabl5547 (18 March 2022), DOI 10.1126/sciadv.abl5547. Retrieved 29 September 2026 from the official Europe PMC supplementary-files endpoint, ZIP member sciadv.abl5547_sm.pdf.
Archive: papers/consciousness_connectomics/tasserie2022_thalamic_dbs_supplement.pdf, 31,550,904 bytes, SHA-256 6d576171c4c8a7cf287af218a7427344159c83e4f6b7f4cd55770eeb7e0ea3c7. The containing singular-name _supplement.zip is 26,267,923 bytes, SHA-256 bb27212f268246af940861b07c503233bf7877953f087244a342087829ace7ca. HTTP 200, application/zip; this successful retrieval supersedes earlier direct-access failures.
Exact reading scope
All 32 pages were read: title/contents, complete Extended Materials and Methods and Extended Results (pages 2–9), every caption and all twelve figure images (pages 10–21), all rows of Tables S1–S7 (pages 22–32). The four-page Table S3 and two-page Tables S6 and S7 were read in full. All figure/table pages were visually rendered and inspected. The text extraction has 1,954 lines; some tables contain overlapping hidden text from earlier layout layers, so their visible PDF renderings, rather than those duplicated extraction lines, were used. No raw recordings, code or analyses were independently rerun. There are no additional videos listed within this supplementary PDF.
Extended methods: what the intervention and analyses actually did
Two macaques, N and T, received a right-thalamus clinical Medtronic 3389 electrode with four contacts. Contacts were 1.5 mm long, spaced by 0.5 mm; external electrode diameter was 1.27 mm. Targeting combined preoperative anatomical MRI, neuronavigation, several atlas/landmark checks, and intraoperative MRI. Figure S1 documents registration, reconstruction of lead trajectory and estimated CT/VL contact locations.
Behavioural testing began at least 20 days after implantation, outside the scanner. The investigators explored voltage amplitude and pulse width at fixed 130 Hz monopolar stimulation, successively applying each contact. High CT stimulation was chosen just above the voltage threshold producing a significant behavioural response. Low CT was below the voltage producing an arousal pattern. The same settings were then used at the VL comparison contact. Thus, high CT is an individually behaviour-titrated intervention; it is not a randomly selected, prospectively fixed population dose. This is useful for demonstrating an intervention under effective settings but limits claims about unselected dose efficacy.
Impedances were checked, delivered current was inspected with an oscilloscope before/after sessions, and stimulation artefacts supplied a benchmark during fMRI. Figures S2–S4 illustrate MRI/DBS artefact suppression and changes in EEG during and after stimulation. Those examples document the processing problem and representative waveforms; they do not independently validate every corrected epoch.
Static analyses used regional time-series correlations, Fisher transformation and run-averaged matrices. Pairwise region tests used t tests and FDR correction. Dynamic analyses vectorized 3,321 connections among 82 regions, clustered the covariance matrices into seven states, and ordered these states by correlation with a structural matrix. Each run's state sequence had 464 entries. The ordering is an analyst-defined similarity rank, not an independently observed level of consciousness.
The structural comparator in Figure S7 is the CoCoMac macaque cortical connectivity atlas, with four ordinal connection-strength grades. It is neither these animals' individual synaptic connectome nor a dynamic measurement. The study asks how functional covariance configurations depart from a fixed atlas template across interventions and states.
Task analyses convolved stimulus categories with a MION haemodynamic response function and derivative; motion and heart rate were nuisance regressors. Global-standard trials immediately following a global deviant were excluded. Extracted beta weights were plotted as percent whole-brain signal relative to implicit rest. Figure S9 spells out the local/global design: five-tone trials, 50-ms tones, 150-ms onset spacing, 850-ms intersequence interval; after habituation, 80% frequent versus 20% rare trial sequences. Local deviance and sequence-level global rarity are separately contrasted.
Extended results and qualifications
Dynamic repertoire. With clustering restricted to awake, anaesthesia and high CT conditions, the most structure-similar state occurred with probabilities .24, .58 and .26, respectively. The probability difference between awake and high CT was nonsignificant (p=.74) and the reported BF01=4.18 favoured the null under the authors' model. Similarity-probability regression slopes were lower in awake/high CT than anaesthesia, with awake versus high CT p=.42, BF01=3.28. These are moderate model-dependent support for similarity on particular metrics, not proof that all neural functions or subjective experience were restored.
For the separate VL clustering, state-7 probabilities were .19 awake, .60 anaesthesia and .72 high VL. Anaesthesia versus high VL provided inconclusive evidence (p=.15; BF10=.64, BF01=1.54), not a demonstrated equivalence. High CT shortened residence in the most structure-like state relative to anaesthesia; high VL retained long residence.
The all-condition analysis is an important nuance. State-7 probabilities were .20 awake, .54 anaesthesia, .37 low CT, .23 high CT, .38 low VL and .63 high VL. Low stimulation at either site produced intermediate changes in this metric. High CT's overall pattern most closely approached wakefulness; it is inaccurate to claim every connectivity marker changes only at CT or that every low-stimulation condition has exactly no effect. Clustering separately in subsets and jointly yields distinct centroids, so state numbers across those fits should not be treated as identical physical states.
Spectral specificity. Figure S5 shows effects on delta, theta, alpha and median power frequency. High VL also reduces delta and elevates median frequency compared with anaesthesia. The consciousness-relevant interpretation therefore depends on the combined behavioural, complexity, connectivity and auditory evidence described in the main paper, rather than a single EEG frequency shift.
Spatial scope of electrical action. Table S1 estimates stimulated nuclei using the Lead-DBS macaque toolbox. In both animals, high CT's estimated volume includes VPM, medial/lateral CM, central/lateral/medial MD, VPL, medial VL and intermediodorsal nuclei: nine listed nuclei/subdivisions. It is not selective stimulation of CM neurons. High VL also overlaps medial/lateral CM in monkey T, whereas its estimated distribution differs in N. These are modelled activation volumes, not direct cell-by-cell measurements. Causal attribution is to an intervention at a regional network access point, with spread and fibre recruitment still relevant.
Physiology covaries with stimulation. Table S2 reports anaesthesia-to-high-CT heart-rate changes from 111±9 to 164±16 bpm in N, and 116±17 to 181±19 bpm in T. In T, mean blood pressure changes from 68±9 to 104±13 mmHg, end-tidal CO2 from 37±2 to 43±2 mmHg, and temperature from 38.3±.7 to 39.6±.4 °C. Oxygen saturation remains approximately 97–99%. VL resting-state rows are supplied only for T: low VL HR145±3 and high VL163±14 bpm. This table alone does not justify saying physiological conditions were unchanged. The task model includes a heart-rate covariate; this reading did not rerun analyses to determine whether physiological covariation accounts for any residual fMRI effects. The asymmetric table should also be preserved when describing the VL evidence.
Activation evidence. Table S3 lists widespread high-CT block-design effects, including bilateral cortical and subcortical regions. Low CT has a smaller set and low VL almost none, whereas high VL affects selected sensorimotor/other areas. Its stated threshold is p<.05 FWE, although individual printed rows sometimes label FDR instead; preserve the reported distinctions if using exact coordinates/statistics.
Figures S10–S12 provide individual local/global contrasts for N and T at p<.001 uncorrected. Tables S5–S6 provide the group-level local/global tests at p<.05 FDR corrected. These evidence levels should not be conflated. The global table includes a few significant anaesthesia effects (for example somatosensory, visual and VPL entries), so “global response absent everywhere under anaesthesia” would be too strong. The authors state no region survives the awake>high-CT global-effect comparison; lack of a significant contrast is not comprehensive equality of responses.
Atlas compression. Table S7 lists the full original CIVM-to-revised-CIVM_R regional mapping, merging subdivisions to match fMRI resolution and deleting certain regions. These matrices describe coarse anatomical regions. They do not estimate microscopic neuronal connectivity or preserve every distinction visible in the anatomical atlas.
Internal reporting issues
The extended VL subsection labels one awake-versus-high-VL statistical comparison as “high CT,” despite its VL context. Figures S11–S12 are headed global effect, but parts of their captions call it local; S12's heading correctly identifies monkey T while a later caption sentence says N. These are unresolved label/copyediting inconsistencies, so the visible headings, design context and group tables were kept distinct rather than silently treated as exact agreement. The raw text extraction also contains hidden duplicated table layers that do not appear in the visible PDF.
Implication for the research synthesis
The supplement strengthens the claim that an anatomically targeted intervention can alter a brain's accessible dynamical repertoire while gross anatomy and continued anaesthetic administration remain in place. It also sharpens the limit: the manipulated cause is a broad electrical intervention with physiological changes, not an isolated nucleus, individual synaptic edge or a complete reconstruction of conscious experience. The structural comparison is an atlas-level explanatory reference. Neither structural similarity nor dissimilarity alone is a universal consciousness test; here it is one operational measurement within a converging experimental design.