# Inner Speech Decoding **Entity class:** Method and research result **Domain:** Speech Neuroprostheses / Neural Decoding **Doc Type:** Method Node **Maturity:** Human-translational **Collection:** [[collections/Neurotech|Neurotech]] **Primary Sources:** [[research/Research Inferences Integration Report - 2026-10-23|Research Inferences Integration Report — 2026-10-23]] ## Definition **Inner speech decoding is the recovery of silently imagined words from neural activity, without any attempt to move the mouth or voice.** The defining result is the Stanford and Emory BrainGate2 study published in *Cell* on 14 August 2025 (senior author Frank Willett; lead author Erin Kunz, with Benyamin Meschede-Krasa), which decoded inner speech from motor-cortex arrays in four people with ALS or brainstem stroke. ## Findings - **Shared code, lower gain.** Attempted and imagined speech activate similar motor-cortex patterns; imagined speech produces the fainter version. - **Accuracy:** sentences decoded from a vocabulary of up to 125,000 words with up to 74% accuracy. - **Thought-password gate:** decoding began only after the user silently imagined a chosen phrase ("Chitty Chitty Bang Bang"), recognized with over 98% accuracy; this is the first implemented cognitive access control in a speech BCI. - **Unprompted decoding:** while participants silently counted shapes on a screen, the decoder produced increasing number sequences, demonstrating that task-related inner speech is readable without an instruction to communicate. ## 2026 extensions - **[[wiki/Paradromics|Paradromics]]** reported distinct neural patterns for attempted and imagined speech in its first [[wiki/Connect-One|Connect-One]] participant, who used the wireless [[wiki/Paradromics Connexus|Connexus]] system for real-time conversation (14 September 2026). - **[[wiki/Neuralink|Neuralink]]**'s [[wiki/Neuralink VOICE Trial|VOICE trial]] showed a participant progressing from mimed speech to thinking words and hearing them spoken in a synthesized voice (18 September 2026). ## Forward Capability Inner speech is the first private mental content to become a decodable signal in clinical practice. The password gate shows the governance answer arriving as engineering: an authorization layer under the user's own control, designed by the research team before any statute asked for it. As decoders move from attempted to imagined speech, the user's effort falls, and the bandwidth ceiling becomes a question of signal quality and model prior rather than of motor function. ## Open Questions - Error rates and latency for open-vocabulary inner speech in daily home use. - Whether a gate trained on one phrase holds across months of electrode drift (INF-0020). - Which access-control standard, if any, the first commercial speech BCIs adopt. ## Research Inferences These entries translate the forward-looking register in [[research/Research Inferences|Research Inferences]] into ordinary wiki prose. The tier labels apply to the inference, not automatically to every factual anchor inside it. Proposed IDs above INF-0500 and tier recommendations come from the 2026-10-23 cycle: [[research/Research Inferences Integration Report - 2026-10-23|Research Inferences Integration Report — 2026-10-23]] · [[research/Research Inferences Integration Audit - 2026-10-23|Research Inferences Integration Audit — 2026-10-23]]. - **INF-0083 — Strongly indicated** (recommended for promotion to established in the 2026-10-23 report). Work through 2026 has moved from decoding externally supplied captions toward recovering participants' own inner speech via subject-specific neural-semantic alignment. Inner speech is the first genuinely private content to become instrumentable, and the technique requires no change to the underlying language model. - **INF-0086 — Strongly indicated.** Semantic decoding from non-invasive recordings requires cooperation today, and that requirement is a property of current signal quality rather than a principle. Legal frameworks built on the assumption of required cooperation are therefore built on a moving quantity. - **INF-0386 — Strongly indicated** (recommended for promotion to established). Coupled cognition requires an access-control model specifying who may read which internal state under what conditions. Access control is a solved discipline in computing, and importing it wholesale is the fastest route to safe coupling. - **INF-0007 — Strongly indicated.** Academic speech BCIs reached 62–78 words per minute with implanted electrodes and model-side decoding, which means the recent gains are attributable to the decoder more than the sensor. Decoder improvement is free of surgical risk and compounds across every existing implant already in a head. - **INF-0507 (proposed) — Established.** A silently imagined password gated an inner-speech decoder with over 98% recognition, and the same system decoded unprompted counting during a visual task; cognitive access control now exists as an engineering primitive. ## Relationships - **program:** [[wiki/BrainGate|BrainGate]] (BrainGate2 clinical trial). - **institution:** [[wiki/Stanford University|Stanford University]] (with Emory University). - **method_family:** [[wiki/Semantic Neural Decoding|Semantic Neural Decoding]] · [[wiki/Speech Restoration|Speech Restoration]]. - **extended_by:** [[wiki/Paradromics Connexus|Paradromics Connexus]] (imagined speech, September 2026) · [[wiki/Neuralink VOICE Trial|Neuralink VOICE Trial]] (thought-to-voice, September 2026). - **population:** [[wiki/Amyotrophic Lateral Sclerosis|Amyotrophic Lateral Sclerosis]] · [[wiki/Locked-In Syndrome|Locked-In Syndrome]]. - **governance_consequence:** [[wiki/Neurorights|Neurorights]] · [[wiki/Cognitive Interface Layer|Cognitive Interface Layer]]. - **ecosystem:** [[wiki/Brain-Computer Interfaces|Brain-Computer Interfaces]] · [[collections/Neurotech|Neurotech]]. - **source context:** [[articles/Technologies for Consciousness Mapping and Transfer|Technologies for Consciousness Mapping and Transfer]] (section VII, the read channel). ## Sources - [Study of promising speech-enabling interface offers hope for restoring communication — Stanford Medicine, August 2025](https://med.stanford.edu/news/all-news/2025/08/brain-computer-interface.html) - [Inner speech decoding device for patients with paralysis — Stanford Report, August 2025](https://news.stanford.edu/stories/2025/08/study-inner-speech-decoding-device-patients-paralysis) - [These brain implants speak your mind, even when you don't want to — NPR, 20 August 2025](https://www.gpb.org/news/shots-health-news/2025/08/20/these-brain-implants-speak-your-mind-even-when-you-dont-want) - [Paradromics Achieves Real-Time Speech with Connexus — Business Wire, 14 September 2026](https://www.businesswire.com/news/home/20260914434545/en/Paradromics-Achieves-Real-Time-Speech-with-Connexus-Brain-Computer-Interface) As of 23 October 2026.