ARCHIVE COPY — SCANNED PREPRINT (ASCII EXPORT) ===================================================== Journal of Cognitive Systems (fictional) 2014; 8(3):115–139 DOI: 10.5555/jcs.2014.8.3.115 Received: 12 Feb 2013 Accepted: 18 Nov 2013 Noninvasive Machine Detection of Exogenous Thought Transmission via Stochastic Resonance in 1/f Noise ---------------------------------------------------------------------------------------------------- Authors ------- A. Voss* (at time of study: Stanford Neuroscience Program, Dept. of Cognitive Science) M. R—— (Dept. of Electrical Engineering, Stanford University) K. —— (Dept. of Statistics, Stanford University) *Correspondence: a.voss [redacted] — current affiliation unknown Affiliations listed reflect the period of data collection and do not imply endorsement by the institutions. Abstract -------- We report above-chance classification of cognitive category at a distance using an unmanned instrument rack ("machine receiver") that models conscious contents as weak structure riding on broadband 1/f environmental fluctuations. In a four-alternative forced-choice (4AFC) paradigm, a Sender viewed randomized images while a physically isolated machine receiver, housed in a separate shielded room with no human present, ran a stochastic-resonance detector tuned to low-SNR pink-noise dynamics. The detector's adaptive phase-lock loop was stabilized by a synthetic self-noise template ("Σ_inst") derived from the instrument's baseline. Across N=240 trials (6 sessions; 40 trials/session), classifier output predicted the Sender's category with 34.6% accuracy (chance=25%; Binomial p=1.2×10^-3; Cohen's h=0.21). Control sessions with no Sender and with desynchronized clocks yielded chance performance (24.9% and 25.3%). Results suggest a weak, nonlocal coupling or uncharacterized conventional channel; replication under stricter isolation is warranted. Keywords -------- consciousness; machine receiver; 1/f noise; stochastic resonance; phase lock; nonlocal coupling Introduction ------------ If conscious contents are stabilized by phase-locking to a broadband 1/f background field, then—under certain assumptions—structure correlated with those contents might be detectable in ambient fluctuations without direct access to the brain. Prior work using a human “receiver” seated next to the instrument (Voss & R——, 2011, tech report) raised confounds. Here we remove the human entirely: only a Sender views stimuli; an unmanned machine receiver, in a separate room, attempts to classify category in real time. Methods ------- Participants: 6 Senders (age 19–31; 3F). No participant was ever present in the receiver room during trials. Task: 4AFC imagery. On each 6‑s trial, the Sender viewed one stimulus drawn from Faces/Objects/Places/Words, randomized without replacement per block. No feedback. Rooms: The Sender room and the Receiver room were 32 m apart in separate corridors with independent HVAC. Receiver room door remained closed. The receiver was an instrument rack; no chair, no workstation login. All networking disabled; clocks synced at session start then allowed to drift per control conditions. Machine Receiver: A software detector ran on an offline workstation, sampling environmental sensors (microphones, photodiodes, accelerometer package mounted to rack) at modest rates. Signals were combined into a pink-noise estimator. An adaptive phase‑lock loop (PLL) was stabilized by Σ_inst, a low‑dimensional vector learned from the instrument’s own baseline self‑noise (3‑min calibration). Classifier: multinomial logistic regression trained on pilot blocks (cross‑validated). No biological signals were recorded. Controls: (a) Sham blocks (no Sender in building); (b) clock offsets (+/‑ 5–10 s); (c) permutation labels; (d) unplugged microphone and photodiode subsets; (e) experimenter‑blind labeling; (f) Faraday cage test for the workstation chassis (room‑level shielding remained partial). Primary outcome: top‑1 accuracy vs. 25% chance. Secondary: AUROC against permuted labels. Results ------- Accuracy: 83/240 correct (34.6%), Binomial p=0.0012, 95% CI [0.29, 0.41]. Sham: 24.9% (p=0.92). Clock offset: 25.3% (ns). Held‑out blocks: 33.8%. Receiver room sensor sub‑ablation (mic only, photo only) reduced effect sizes but did not abolish above‑chance trends. Permutation tests (10,000) yielded p=0.0021. Artifacts considered: air handling correlation (mitigated via independent HVAC), building vibration (accelerometer features down‑weighted with regularization), RF ingress (room not a full Faraday enclosure). Limitations ----------- Effect sizes are small and parameter‑sensitive. The machine receiver classifies coarse category only; it does not read thoughts. Partial shielding leaves open conventional leakage routes. Results require preregistered replication under stronger isolation. Discussion ---------- Removing the human “receiver” reduces a key confound and still yields modest above‑chance classification, consistent with—but not proof of—nonlocal coupling models. Alternatively, a subtle conventional pathway may remain. If a coupling exists, it appears stochastic and best detected via resonance techniques rather than direct energy transfer. Code & Data Availability ------------------------ Receiver code (v0.6‑m) archived on internal Stanford server at submission; Σ_inst synthesis module withheld pending IP review. De‑identified machine logs available upon request (redacted). Acknowledgments --------------- We thank lab volunteers and two anonymous reviewers for critiques that improved controls. Early prototype work was supported in part by a small private grant (details redacted). No conflicts declared. Peer Review Notes (excerpt) --------------------------- Reviewer A: “Effect size is modest; unmanned design is a genuine improvement. Recommend accept with expanded permutation tests and stronger RF controls.” Reviewer B: “Remove ‘telepathy’ language; frame as weak nonlocal correlation. Suggest full Faraday enclosure in replication.” Citation -------- Voss A, R—— M, —— K. Noninvasive machine detection of exogenous thought transmission via stochastic resonance in 1/f noise. J Cogn Syst. 2014;8(3):115–139. doi:10.5555/jcs.2014.8.3.115 Editorial Note (2016) --------------------- Following post‑publication discussion, the journal issued an Expression of Concern requesting larger‑scale replication and stricter shielding. The authors reported lab dissolution and data archiving. No retraction was issued.