Research Library
A searchable portal connecting papers, books, themes, DOIs, lectures and novel constructs—with a summary of what each source actually supports.
This is a curated working library, not a complete systematic review. DOI links resolve to the publisher or index record.
65 research anchors
Reframes drug-produced internal states as learned stimuli that can predict outcomes, signal which response will work or modulate what another cue means. It separates these stimulus functions from reinforcement while tracing candidate body–brain routes and future treatment questions.
A three-stage account of binge/intoxication, withdrawal/negative affect and preoccupation/anticipation that anchors the addiction arc without reducing recovery to circuitry alone.
A major synthesis of reward, stress and executive-control changes. Flow Hijacked uses it as one layer of the field, not a total description of the person.
A randomized trial comparing maintenance approaches. It supports learning to contact difficult states earlier while keeping claims tied to the studied intervention and population.
Distinguishes sensitized wanting from pleasure. This helps explain why an old route can regain force even when a person no longer endorses or enjoys it.
Tracks shifts from goal-directed action toward habitual and compulsive control, while leaving room for heterogeneity and multiple routes into persistent use.
Places prediction inside an embodied, situated agent. It supports the project’s insistence that brain, body and world cannot be cleanly separated in a living trajectory.
Shows why slowing recovery and rising variability can precede regime shifts in complex systems. Clinical transfer remains a hypothesis requiring direct validation.
Connects dynamical early-warning ideas with intensive mood data. It is an important bridge, but not a validated relapse detector for an individual person.
A foundational account of autonomy, competence and relatedness. These needs help describe why a technically abstinent life may still remain motivationally narrow.
Moves wellbeing beyond momentary happiness toward purpose, growth, autonomy, relationships and environmental mastery—useful coordinates for a wider life.
Treats reliable social proximity as part of the expected energetic baseline. It supports belonging as a control resource without claiming that relationships erase risk.
Frames recovery partly as movement between social identities and groups. The field changes when sober belonging becomes credible, enacted and reinforced.
A bridge from psychiatric phenomena to network dynamics and state transitions. It supports dynamical formulation while preserving the gap between model and clinical instrument.
The principal clinical lineage for mortality, freedom, isolation and meaninglessness in Lecture 36.
A predictive-processing account linking interoceptive inference to emotion and embodied selfhood. It supports one influential model of experience, not a settled solution to consciousness.
TMS–EEG showed that a cortical perturbation becomes local and stereotyped during non-REM sleep compared with wakefulness. It is part of the empirical lineage behind perturb-and-record approaches to conscious state.
Introduces the Perturbational Complexity Index from the compressibility of TMS-evoked EEG responses. PCI distinguishes studied conscious and unconscious conditions; it is not a complete theory or measure of a person’s inner richness.
A foundational process-based account arguing for testable change processes and idiographic functional models rather than protocol labels alone. It supports process and route comparison without validating Flow Hijacked’s TVFR proposal itself.
A large meta-analysis finding a robust association between therapeutic alliance and psychotherapy outcome across approaches and settings. It makes alliance clinically consequential without establishing that alliance alone causes every outcome.
Translates inhibitory-learning research into exposure strategies centered on expectancy violation, variability, retrieval and generalization. It directly supports controlled perturbation in exposure-based work without generalizing that mechanism to every psychotherapy.
A synthesis of how dopamine responses can encode differences between expected and obtained outcomes while remaining heterogeneous across components and conditions. It anchors the learning claim without turning dopamine into a pleasure meter.
Shows how modulatory substances reconfigure circuit output rather than assign one fixed function to one molecule. It supports the lecture’s move from isolated chemical levels to receptor-, circuit- and state-dependent effects.
A foundational adaptive-gain account of locus-coeruleus function linking arousal, task engagement and behavioral flexibility. It supports state-dependent gain while cautioning against a simple more-is-better model.
Reviews glutamatergic, GABAergic and connectivity changes in depression and their treatment relevance. It supports a plasticity-and-network account rather than a one-transmitter diagnostic story.
Separates reward-learning and anhedonia processes within depression and connects them to stress. It supports asking which component of anticipation, effort, learning or pleasure has changed instead of inferring one low chemical level.
Finds substantially lower mortality while people are retained in methadone or buprenorphine treatment, with heightened risk around treatment entry and discontinuation. It anchors medication as a safety intervention, not a moral shortcut.
A large network meta-analysis supporting average antidepressant efficacy while showing meaningful differences in efficacy and acceptability. Population averages inform care but do not predict one person’s response or replace individualized assessment.
Estimates that 9.1% of people with major depressive disorder globally received minimally adequate treatment in 2021. The denominator is people with MDD—not the world population—and wide data gaps make this a modeled service-coverage estimate rather than a census.
Identified 1,030 symptom profiles among 3,703 STAR*D participants, making within-diagnosis heterogeneity impossible to treat as noise. The result supports component-level formulation; it does not invalidate diagnosis or create a validated personalized treatment selector.
The landmark PET study linked shifts in negative mood to reciprocal changes involving subgenual cingulate and right prefrontal regions. It established an influential circuit hypothesis around area 25, not a single-region cause or individual diagnostic scan.
A small open-label proof-of-concept study of subgenual cingulate white-matter DBS reported sustained improvement in some people with severe treatment-resistant depression. Its historical importance is substantial, but its sample and uncontrolled design cannot establish efficacy by themselves.
BROADEN was stopped after a futility analysis: active SCC-DBS did not significantly separate from sham during the six-month blinded phase (20% versus 17% response), and serious procedure- or device-related events occurred. Any later DBS optimism must remain in dialogue with this controlled negative result.
Across 1,801 patients and controls, multivariate neuroimaging models reached only modest out-of-sample diagnostic performance; the best accuracy was about 62%. The study is a useful boundary for computational psychiatry: group signal is not yet an individual clinical biomarker.
A network meta-analysis of 331 randomized trials found that several major psychotherapies outperform usual-care or waiting controls, with many head-to-head differences small or uncertain. Modality, alliance, access, preference and the person’s maintaining processes still matter.
Across 218 randomized trials, exercise reduced depressive symptoms, with walking or jogging, strength training and yoga among the supported options. Confidence for specific modality rankings was low or very low, and disability, illness severity, safety, access and adherence constrain any prescription.
The pooled signal suggested a possible small-to-moderate benefit, but average effects were not clinically important, bias may have favored creatine and the evidence was very uncertain. Creatine remains a candidate adjunct for research and clinician review, not a stand-alone depression treatment.
Adverse experiences were reported across meditation studies, with estimates varying sharply by design and reporting method. The review argues for screening, informed choice, monitoring and routes to stop or obtain help—not for treating meditation as universally harmful or universally safe.
Reviews conceptual and empirical problems in translating critical-slowing-down signals into psychopathology. Early-warning dynamics remain a research program, not a validated personal relapse detector, and false alarms or missed transitions can carry clinical costs.
In a large cohort comparison, a collaborative Safety Planning Intervention plus structured follow-up was associated with fewer suicidal behaviors and greater outpatient engagement than usual care. It supports concrete plans and human continuity; it does not make any risk score perfectly predictive.
The updated Cochrane review found low- to very-low-certainty evidence and a small-to-modest placebo difference unlikely to be meaningful for most adults with MDD. Omega-3 should not be presented as a proven substitute for assessment or established care.
Randomized dietary interventions showed a modest average reduction in depressive symptoms, while anxiety effects were not clear. Many samples were not recruited for clinical depression and mood was often a secondary outcome, so food belongs in whole-person care without becoming a cure narrative.
Pooled follow-up across five SCC-DBS studies reported sustained improvement among 172 implanted patients, with primary analyses through 24 months. Pooling, attrition and largely uncontrolled long-term observation limit causal inference; these data cannot erase BROADEN’s negative blinded phase or the demands of invasive specialist care.
In a second double-blind trial, 48 randomized participants received active or sham SNT; one-month remission was 50.0% versus 20.8%. The replication strengthens the signal, while the small specialized sample, short primary horizon, intensive schedule and need for durability and external-effectiveness data keep implementation conditional.
RECOVER randomized 493 adults with marked treatment-resistant depression. Active VNS did not significantly outperform sham on the primary time-in-response outcome, although several secondary symptom and function measures favored active treatment; interpretation must preserve both findings and the realities of surgery, programming and long-term monitoring.
CREST–MST found MST non-inferior to modern right-unilateral ultra-brief ECT within its prespecified margin for remission, with less autobiographical-memory worsening. MST still induces a therapeutic seizure and requires anesthesia, seizure-service governance, trained teams and regulatory implementation; it is not minimal-intervention self-care.
In rats, disrupting protein synthesis in the amygdala after retrieval impaired a consolidated auditory fear memory. The post-reactivation asymmetry became a foundational causal demonstration of reconsolidation; it does not show that ordinary human remembering automatically rewrites memory.
Mouse contextual-fear experiments separated reminder conditions associated with reconsolidation from longer exposure associated with extinction, with different time and biochemical dependencies. The result supports a process boundary, but the exact boundary moves with task, memory and protocol.
In a crab avoidance paradigm, mismatch between expected and actual reminder events helped determine whether reactivated learning entered reconsolidation or extinction. It is a key prediction-error lineage, although its operational details cannot be transferred wholesale to human autobiographical or clinical memory.
Healthy adults who received a subtle contextual reminder before learning a second list later incorporated more new items into recall of the first list. The study supports reminder-dependent interaction between episodes; it does not by itself identify a molecular reconsolidation mechanism.
In rats, extinction delivered after a brief retrieval trial within a proposed reconsolidation window produced unusually persistent attenuation of conditioned fear. The finding launched a major translational program, but human reliability and mechanism remain contested.
A human conditioning study reported reduced return of fear when extinction followed reminder within a proposed reconsolidation window. It is conceptually important, but later close and registered replications have not established a robust general effect.
Human fear retrieval did not become pharmacologically disruptable without a relevant expectancy violation. The study directly challenges the idea that retrieval is sufficient and makes prediction error a candidate gate rather than a decorative synonym for surprise.
A foundational associative-learning synthesis explains extinction as context-sensitive new learning rather than simple erasure. Renewal, reinstatement and spontaneous recovery show that old learning can remain available and regain control under changed retrieval conditions.
A Bayesian latent-cause model formalizes how learners infer whether new observations belong to an old hidden cause or a new one. It explains context-sensitive extinction and relapse patterns computationally; latent causes are inferred model variables, not directly observed neural objects.
Computational analyses connect memory modification to inference over hidden causes: reminder timing and gradual versus abrupt extinction can change whether new evidence is assigned to old or new learning. The model generates testable predictions but is not proof of one biological implementation.
A preregistered human study failed to reproduce the expected retrieval-extinction boundary pattern. Protocol sensitivity may still matter, but the negative result is strong evidence against presenting the effect as reliable, simple or ready for routine clinical translation.
A contemporary cross-species synthesis argues that destabilization depends on more than mismatch magnitude: prior learning history, reminder structure and memory features shape lability. It strengthens a conditional account while warning against a universal prediction-error dose curve.
Rat experiments support a non-monotonic assignment problem: prediction-error conditions influenced whether extinction appeared to update old fear learning or create a separate memory. The result directly informs ARC, but one paradigm does not establish a universal three-zone law.
A systematic review and meta-analysis found insufficient convincing evidence for propranolol as a PTSD treatment through presumed reconsolidation disruption. Heterogeneous protocols and small studies leave questions open, but clinical claims must remain well below the certainty of the animal mechanism literature.