The DOI matches its Crossref record; 212 of the 212 references it cites were found in a registry or an index; 187 of the 252 sources its claims cite were checked for support.
The speech neuroprosthesis
Open accessAbstract
Review of advancements in speech neuroprostheses decoding speech production directly from motor-planning and vocal-tract articulatory cortical representations, highlighting high-resolution neural interfaces and deep learning speech computational algorithms.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
55 Partly supported · 35 Own interpretation · 26 Supported · 17 No full text of the sources · 13 Sources not found · 4 No basis given · 3 Supported so far · 3 Couldn’t confirm · 3 Not checked yet · 1 Mixed support · 1 Not supported · accuracy not yet measured
Some cited sources support this claim and some don’t — read each source’s verdict below.
In the paper
“Another approach, pursued by Angrick et al.111 and Anumanchipalli et al.110, involved a two-stage decoding process. In the first stage, a deep-learning model is trained to regress neural activity (recorded from SEEG111 or ECoG110) to a time series of acoustic features, such as the mel-spectrogram (Fig. 3b). In the second stage, a speech synthesizer — a vocoder — is used to convert the acoustic rep”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Partly supportedcites [111] sourceExperimental — accuracy not yet measured
The Decoding Approach section describes the claimed two-stage process: a DenseNet maps neural activity to mel-scaled spectrograms, and a WaveNet vocoder produces audio. However, the citing sentence assigns SEEG to Angrick et al.; the Experiment and Data Recording section says they recorded ECoG using a cortical electrode grid. Page numbers are unavailable in our copy. (LLM, conf=0.90)
- Disputedcites [110] sourceExperimental — accuracy not yet measured
The Speech decoder design and Methods (Neural decoder; page numbers unavailable) describe a different two-stage process: an ECoG-based neural network first decodes articulatory features, then another network converts those features to acoustic features. Audio is synthesized afterward, not as the second decoding stage. The paper treats direct neural-to-acoustic decoding as a separate comparator. (LLM, conf=0.89)
Checked by a reasoning model — may be wrong.
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
At least one cited source doesn’t support this claim or points to a different paper, and none confirmed it — see the verdicts below.
In the paper
“Together, imagined, silently attempted and attempted speech reflect intended speech from the participant, which is distinct from internal monologues or thoughts139.”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Evaluation and standardization
What the sources say
- Disputedcites [139] sourceExperimental — accuracy not yet measured
In “Speech Neuroprosthesis as a Mind-Reading Device” (page unavailable), the paper says covert speech can include internal monologue and that distinguishing speech intended for externalisation from speech not intended for it is a central difficulty. It also warns that imagined speech could be captured without the user intending to express it. This contradicts the claim that imagined speech is necessarily intended speech distinct from thoughts or internal monologue. (LLM, conf=0.86)
Checked by a reasoning model — may be wrong.
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Disputed
Every source the paper cites for this claim was checked and supports it.
In the paper
“Studies furthering our basic understanding of the cortical encoding of speech features, notably motor control of the vocal-tract articulators during speech, laid the groundwork for a speech neuroprosthesis that decoded text in the form of words and sentences from the cortical activity of an individual with incomplete locked-in syndrome who relied on AAC methods to communicate12.”
§ The speech neuroprosthesis > Introduction
What the sources say
- Supportedcites [12] source
The Discussion reports real-time decoding of full words and sentences from the participant’s cortical activity and connects the result to earlier work on speech-production representations. The Methods (Participant) describe severe paralysis and anarthria after a pontine stroke, with residual head movement used to operate an assistive typing interface—consistent with the citing sentence’s description of incomplete locked-in syndrome and reliance on AAC. (LLM, conf=0.86)
plaindirectSingle case- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Bouchard et al.17 first demonstrated the dynamics of somatotopic speech-articulator representations in the SMC for the jaw, lips, tongue and larynx during speech production of syllables (Fig. 1; Bouchard et al. 2013).”
§ The speech neuroprosthesis > Cortical encoding of speech features
What the sources say
- Supportedcites [17] source
The Abstract reports cortical recordings during consonant-vowel syllable production and says the articulator representations were somatotopically arranged and temporally coordinated. The Spatial Representation of Articulators and Timing of Articulator Representations sections identify the jaw, lips, tongue, and larynx and describe their spatial arrangement and timing during syllable production. (LLM, conf=0.92)
boosted- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Neural populations in the ventral sensorimotor cortex (vSMC) and the middle precentral gyrus (midPrCG) control the millisecond-level movements of the vocal-tract articulators (lips, tongue, jaw and larynx) that give rise to speech sounds17,18 (Fig. 2a).”
§ The speech neuroprosthesis > Overview of speech features
What the sources say
- Partly supportedcites [17] sourceExperimental — accuracy not yet measured
The Abstract and Discussion describe vSMC representations of vocal-tract articulators and their coordination in producing speech sounds. The Timing of Articulator Representations section places articulator movements on a scale of tens of milliseconds. The paper does not identify the middle precentral gyrus (midPrCG) as a separate region controlling those movements; page numbers are unavailable in our copy. (LLM, conf=0.86)
- Partly supportedcites [18] sourceExperimental — accuracy not yet measured
The Abstract (page unavailable) reports somatomotor maps of vocal-tract articulation spanning primary motor and somatosensory areas, and the Results locate articulation-related activity on the precentral gyrus. But the paper does not identify the middle precentral gyrus specifically or establish millisecond-level control by these neural populations. The Discussion (page unavailable) says its findings do not directly address the planning and sequencing of articulator movements. (LLM, conf=0.86)
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Interestingly, at the neuronal level, locations on the anterior precentral gyrus in orofacial or hand areas contain single neurons tuned to movements of the whole body, including key speech articulators15,51,52.”
§ The speech neuroprosthesis > Cortical encoding of speech features
What the sources say
- Partly supportedcites [15] sourceExperimental — accuracy not yet measured
The Main section (page not provided) reports intermixed tuning to the jaw, larynx, lips and tongue at a single-neuron level in area 6v. It studied attempted orofacial movements and speech, but does not establish tuning to movements of the whole body or report such tuning in hand areas. (LLM, conf=0.88)
- Supportedcites [51] source
The Results section reports tuning to face movements, including mouth, tongue and speaking, alongside head, arm and leg movements in the hand-knob area of the precentral gyrus. It also reports that well-isolated single neurons were frequently tuned to multiple movement types. This paper supports the hand-area portion of the sentence; the sentence groups it with other citations for orofacial areas. (LLM, conf=0.90)
- Partly supportedcites [52] sourceExperimental — accuracy not yet measured
The Discussion reports neurons responding to mouth and tongue movements in the dorsal arm-and-hand area of motor cortex. It describes representation of much of the body as a suggestion based partly on emerging work, not a finding that individual neurons are tuned to whole-body movements. This study recorded in the hand area of two people with tetraplegia, not in orofacial areas. (LLM, conf=0.88)
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Current speech-decoding applications from brain activity are primarily aimed at restoring communication to individuals with both limb and vocal-tract paralysis14–16.”
§ The speech neuroprosthesis > Neurological speech disorders impairing communication
What the sources say
- Supportedcites [14] source
The Abstract describes speech neuroprostheses as a way to restore communication to people with paralysis and identifies this study’s participant as having severe limb and vocal paralysis. The Overview likewise says the system was designed to enable that participant to communicate by decoding intended speech. (LLM, conf=0.88)
- Partly supportedcites [15] sourceExperimental — accuracy not yet measured
The Abstract and Discussion describe speech BCIs as a way to restore communication to people with paralysis and discuss people with severe speech and motor impairment. They do not establish that current applications are primarily aimed at people with both limb and vocal-tract paralysis. The Main section says this participant retained limited orofacial movement and could vocalize, though she could not speak intelligibly. (LLM, conf=0.84)
- Partly supportedcites [16] sourceExperimental — accuracy not yet measured
The Introduction describes speech brain-computer interfaces as a way to restore lost communication. But the Methods (Study participant) describe a person with severe limb impairment and dysarthria, and the Limitations (Decoding speech in patients with other neurological impairments) state that he retained voluntary bulbar muscle control. The paper therefore does not establish that current applications are primarily aimed at people with both limb and vocal-tract paralysis. (LLM, conf=0.87)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The sources with a verdict support this claim; the rest weren’t checked yet, have no full text we could read, or their verdicts aren’t rated yet.
In the paper
“Years of work focused on mapping premotor and primary motor representations on the precentral gyrus and central sulcus, revealing a somatotopic organization in which neural control of muscle groups is represented along the ventral–dorsal axis34–38.”
§ The speech neuroprosthesis > Cortical encoding of speech features
What the sources say
- Not checked — no full text of the sourcecites [34] source
- Not matchedcites [35]
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Not checked — no full text of the sourcecites [36] source
- Not checked — no full text of the sourcecites [37] source
- Supportedcites [38] source
The Main section describes the long-established somatotopic map of primary motor cortex and says later imaging and electrophysiology recapitulated it. It also locates foot representation near the midline, hand superior-laterally, and tongue laterally, consistent with the cited sentence’s broad description of an anatomical ordering of body-part representations. Page numbers are not provided in our copy. (LLM, conf=0.88)
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Subsequent work discovered that SMC populations encode articulatory kinematic trajectories and vocal-tract configurations43 during continuous speech24,39,44,45 (Fig. 1; Mugler et al. 2018, Chartier et al. 2018 and Dichter et al. 2018).”
§ The speech neuroprosthesis > Cortical encoding of speech features
What the sources say
- Partly supportedcites [43] sourceExperimental — accuracy not yet measured
The Discussion reports that precentral-gyrus activity represented articulatory gestures more strongly than phonemes. But the Results describe recordings during single-word, monosyllabic utterances, not continuous speech. The paper estimated articulator trajectories to label gestures; it does not establish that the recorded populations encoded continuous kinematic trajectories or vocal-tract configurations. (LLM, conf=0.87)
- Supportedcites [24] source
The Discussion (page not supplied) states that discrete neural populations in human sensorimotor cortex encode articulatory kinematic trajectories during continuous speech production. The Results section on single vSMC electrodes (page not supplied) also describes trajectories corresponding to specific vocal-tract configurations. (LLM, conf=0.94)
- Partly supportedcites [39] sourceExperimental — accuracy not yet measured
The Volumetric Results and Surface Results (pages not provided) locate somatomotor activity associated with articulation, tongue movement, and vocalization during a short syllable task. They do not show that SMC populations encode vocal-tract configurations or articulatory kinematic trajectories during continuous speech. (LLM, conf=0.78)
- Partly supportedcites [44] sourceExperimental — accuracy not yet measured
The Results section ('dLMC and vLMC encoding of pitch components') reports that sensorimotor-cortex electrodes encode pitch accent, phrase, and voicing during spoken sentences. But the Methods ('Acoustic analysis') explicitly says the model does not track articulators directly or model their movements, so this paper does not establish encoding of articulatory kinematic trajectories. The citing sentence places a separate citation immediately after 'vocal-tract configurations.' (LLM, conf=0.84)
- Could not verifycites [45] sourceExperimental — accuracy not yet measured
The Cerebellum section describes mapping lip, tongue and laryngeal activity during vocalization, and the Motor Cortex section discusses laryngeal representations. Neither reports that SMC populations encode articulatory trajectories or vocal-tract configurations during continuous speech (searched for those terms and related descriptions). The indexed passages provide no page numbers. (LLM, conf=0.75)
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Chartier et al.24 discovered that the SMC encodes a large variety of low-dimensional movement trajectories of the vocal tract (also called gestures), which could be directly related to all consonants and vowels.”
§ The speech neuroprosthesis > Cortical encoding of speech features
What the sources say
- Partly supportedcites [24] sourceExperimental — accuracy not yet measured
The Results (Kinematic organization of vSMC; page unavailable) report distinct articulatory trajectories associated with broad consonant groupings and vowel clusters, supporting the central claim that vSMC encodes vocal-tract movements related to speech sounds. The paper says the consonant groupings broadly cover English consonants, but its analysis excludes liquids and does not establish a direct correspondence to every consonant and vowel. The Methods describe concise seven-value representations of the trajectories, though they do not characterize the full trajectories as low-dimensional. (LLM, conf=0.84)
hedgedcorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Recent studies have also achieved high-performance speech-decoding results15,16 by measuring relevant single-unit and multi-unit neuronal activity from multiple small (4 × 4 mm2 (ref. 83)) cortical regions on the anterior precentral gyrus15,16,52,98.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Neural-recording interfaces
What the sources say
- Partly supportedcites [15] sourceExperimental — accuracy not yet measured
The Main text reports speech-related recordings from two small area 6v arrays and says both contributed to decoding; the Discussion reports decoding unconstrained sentences at 62 words per minute using neural activity recorded at single-neuron resolution. The passages do not establish that the decoder measured multi-unit activity, so that part of the attribution is unsupported. (LLM, conf=0.82)
- Partly supportedcites [16] sourceExperimental — accuracy not yet measured
The Methods (Surgical implantation; Recording array implant locations and decoding contributions, pages not provided) describe four small arrays targeting precentral-gyrus regions, with electrodes designed to record from one or a few neurons. The Results (Online decoding performance, page not provided) report high-accuracy speech decoding. The copy does not establish that all the recorded regions were on the anterior portion of the precentral gyrus; it includes an array targeting primary motor cortex (area 4). (LLM, conf=0.78)
- Partly supportedcites [52] sourceExperimental — accuracy not yet measured
The Results report above-chance offline decoding of spoken syllables and words from intracortical activity, and describe single-unit and threshold-crossing activity recorded from two arrays per participant in the dorsal motor-cortex ‘hand knob’ area. The Neural recording methods likewise place the arrays in that area, but do not establish the claim’s more specific ‘anterior precentral gyrus’ location or array dimensions; the Discussion also limits the result to a small set of sounds in participants able to speak. (LLM, conf=0.84)
- Partly supportedcites [98] sourceExperimental — accuracy not yet measured
The Methods describe two small Utah arrays in the dorsal precentral ‘hand knob’ area, and the Discussion reports speech-decoding performance comparable to ECoG studies. But the Methods say the spikes were not sorted into single units, and the Results say high-frequency local field potentials—not spike activity—gave the better decoding performance. The paper does not establish the claim’s specific anterior location or its characterization of the results as achieved by measuring single- and multi-unit activity. (LLM, conf=0.86)
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Along these lines, when able speakers are instructed to imagine speaking or hearing a word, neural activity in the STG can be used to decode those words with above-chance accuracy59–63.”
§ The speech neuroprosthesis > Cortical encoding of speech features
What the sources say
- Supportedcites [59] source
The Materials and Methods (Experimental paradigm; page unavailable) say participants were instructed to imagine hearing a word. The Results (Anatomical distribution of discriminant electrodes; page unavailable) report significant classification using only STG electrodes at the group level (mean 58%) and in four of five participants. (LLM, conf=0.91)
- Partly supportedcites [60] sourceExperimental — accuracy not yet measured
The Results (§3.1; page unavailable) report above-chance decoding of vowels and consonant pairs in imagined words, and §3.2 identifies temporal speech regions among the informative sites. The study did not demonstrate decoding whole words from STG activity, and its task was covert repetition of visually presented words, not a separate instruction to imagine hearing them. (LLM, conf=0.88)
- Partly supportedcites [61] sourceExperimental — accuracy not yet measured
The Results (Covert speech; page not provided) report above-chance identification of speech segments during silent reading, and the Discussion identifies the STG as an informative region for decoding covert speech features. The study did not demonstrate decoding individual imagined words or separately test instructions to imagine hearing a word; its identification result concerns segments of continuous speech. (LLM, conf=0.88)
- Partly supportedcites [62] sourceExperimental — accuracy not yet measured
The Results (‘Decoding imagined speech’; page unavailable) report above-chance classification in studies whose participants imagined hearing or speaking, but the classifier distinguished groups of words rather than decoding individual words. The Results (‘Speech task and item discrimination’; page unavailable) find item-related signals in bilateral superior temporal cortex, but do not establish that STG activity alone decoded the words above chance. (LLM, conf=0.86)
- Partly supportedcites [63] sourceExperimental — accuracy not yet measured
The Searchlight Results (p. 5) identify STG voxels that distinguish semantic categories when participants were presented with spoken or written words. But the Experimental Design (p. 3) describes covert recall of category exemplars, not instructions to imagine speaking or hearing a word; the paper reports decoding animal-versus-tool categories, not individual imagined words. (LLM, conf=0.88)
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“In able speakers, single-word classification has been highly successful using predefined and restricted vocabularies115,116.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Supportedcites [115] source
The Stimuli and task section says participants read aloud words from a predefined set of twelve. The Machine learning classifiers section reports 92–100% accuracy in classifying those words, supporting the claim for this restricted vocabulary. Page numbers were not provided for these passages. (LLM, conf=0.93)
- Partly supportedcites [116] sourceExperimental — accuracy not yet measured
The “Decoding inner speech using electrocorticography” section reports significantly above-chance classification of individual words across five patients, with a mean accuracy of 69% for word pairs versus 50% chance. The “Opportunities” section describes this as decoding a limited vocabulary. Calling it “highly successful” overstates what the paper presents as a proof of concept, and the cited passages do not establish that the patients were able speakers (page numbers unavailable in our copy). (LLM, conf=0.83)
boostedcorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Furthermore, when an individual with anarthria silently attempted to speak — that is, was asked to attempt to ‘mime’ with no audible output from that attempt — select neural populations in the STG contributed to decoding performance14.”
§ The speech neuroprosthesis > Cortical encoding of speech features
What the sources say
- Partly supportedcites [14] sourceExperimental — accuracy not yet measured
The Overview of multimodal speech-decoding system says the participant had anarthria, silently attempted speech without vocalizing, and had electrodes over the superior temporal gyrus (STG). Articulatory representations drive decoding says temporal-lobe electrodes contributed to decoding, but cautions that they may have recorded activity from the sub-central gyrus rather than temporal-lobe sites. It therefore does not establish that neural populations in the STG itself contributed. (LLM, conf=0.82)
plaincorrelationalSingle case- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“To build on this result and move beyond direct classification of words or sentences, Makin et al.108 applied a recurrent neural network (RNN) encoder-decoder framework to encode temporal patterns in neural activity during sentence production into an abstract representation that was then decoded word-by-word into phrases (Fig. 1; Makin et al. 2020).”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Supportedcites [108] source
The Results (Decoding pipeline; page not specified) describe an encoder RNN that turns the neural sequence into a single representation. The Methods (The network; page not specified) describe a decoder RNN initialized with that representation and predicting words sequentially from neural activity recorded during sentence production. (LLM, conf=0.93)
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Intracortical MEA recordings are used to measure neural activity at the level of single neurons or neuron clusters83.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Neural-recording interfaces
What the sources say
- Supportedcites [83] source
The paper’s introductory passage under “Classification of Evidence” (page not provided) says intracortical BCI sensors can access signals from single neurons and local field potentials. Its Methods figure identifies the recording sensor as an intracortical microelectrode array, supporting the claim about MEA recordings. (LLM, conf=0.88)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Semantics appear to be encoded in a distributed manner, across many cortical regions including prefrontal, speech and language association cortices67,68 (Fig. 1; Tang et al. 2023).”
§ The speech neuroprosthesis > Cortical encoding of speech features
What the sources say
- Supportedcites [67] sourceExperimental — accuracy not yet measured
In “Mapping semantic representation across cortex,” the paper describes semantic information as distributed across broad prefrontal, temporal, and parietal regions. Its semantic-atlas section also identifies areas in lateral temporal cortex, which it discusses in relation to language comprehension. These passages support the claim’s broad description of distributed semantic encoding, including prefrontal and language-associated cortex. (LLM, conf=0.89)
- Supportedcites [68] sourceExperimental — accuracy not yet measured
The Results section (‘Decoding across cortical regions’) reports above-chance decoding of words and phrases from the prefrontal, association, and speech regions, and finds overlapping word-level representations across regions. It also notes that semantic decoding from the speech network was less consistent than from the other two regions. (LLM, conf=0.89)
hedgedcorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The sources with a verdict support this claim; the rest weren’t checked yet, have no full text we could read, or their verdicts aren’t rated yet.
In the paper
“Despite this, durable stability has enabled ECoG-based neuroprostheses without the need for daily calibration by the user12,14,94,95.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Neural-recording interfaces
What the sources say
- Supportedcites [12] source
The Results section (‘Long-term stability of acquired cortical signals’) reports consistent model performance throughout the 81-week study without daily or weekly recalibration. The Discussion attributes the ability to accumulate training data without daily recalibration to the stability of ECoG recordings. (LLM, conf=0.93)
- Supportedcites [14] source
The Text decoding section reports high classification accuracy across 19 sessions after the classifier was frozen, including after a 61-day recording hiatus, and explicitly says this demonstrates stability without recalibration. The Discussion likewise attributes across-day testing without same-day retraining to stable cortical-surface recordings. (LLM, conf=0.92)
- Not checked — no full text of the sourcecites [94] source
- Supportedcites [95] source
The Results section, “Stable Decoder Performance Over Three Months” (page unavailable), reports that the ECoG speech BCI was tested for three months with no separate baseline collection or model recalibration before real-time sessions. The Discussion (page unavailable) likewise says the study eliminated model retraining and baseline recalibration. (LLM, conf=0.93)
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Importantly, those demonstrations in able speakers with epilepsy served as a proof-of-principle for future applications in individuals with vocal-tract paralysis that led to the first clinical trial for a speech neuroprosthesis using a chronically implanted 128-electrode ECoG array12.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Neural-recording interfaces
What the sources say
- Partly supportedcites [12] sourceExperimental — accuracy not yet measured
Split into 2 atoms — 0 confirmed, 2 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Furthermore, a 253-electrode ECoG array with increased channel count and density led to substantial gains towards high-performance speech decoding14.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Neural-recording interfaces
What the sources say
- Supportedcites [14] source
The Overview describes a 253-channel high-density ECoG array used for speech decoding. The ‘Articulatory representations drive decoding’ section reports worse performance when lower-density sampling was simulated, and the Discussion identifies denser cortical sampling as one of several innovations that enabled the performance gains. The paper does not attribute those gains to the array alone. (LLM, conf=0.90)
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Three studies (each in a different individual with vocal-tract paralysis: two with MEAs15,16 and one with ECoG14) leveraged RNN models trained to map an input sequence of neural activity to an output sequence of phonemes without the need for alignment between the neural activity and ground-truth sequence of phonemes”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Supportedcites [15] source
The Main section describes a participant with ALS who could no longer speak intelligibly and whose neural activity was recorded with microelectrode arrays. In ‘Decoding attempted speech’ (Fig. 2a; page unavailable), the paper describes an RNN that maps neural-activity time series to phoneme probabilities using a connectionist temporal classification training procedure. This supports the paper’s role as one of the cited studies; it does not need to document the other two. (LLM, conf=0.90)
- Partly supportedcites [16] sourceExperimental — accuracy not yet measured
The Methods (‘Decoding speech’; page unavailable) describe predicting phonemes from cortical neural activity during attempted speech, and the Introduction identifies the participant as having advanced ALS and severe dysarthria. Our copy does not establish that this study used an RNN trained with CTC loss or decoded without phoneme alignment; it also describes impaired speech, not complete vocal-tract paralysis. (LLM, conf=0.82)
- Supportedcites [14] source
The Methods sections on Neural implant and Text decoding identify this as the ECoG study and state that its RNN was trained with CTC loss to predict phone sequences without temporal alignment between neural activity and phone labels. The citation identifies this study as the ECoG example in the sentence; the two MEA studies have separate citations. (LLM, conf=0.93)
plainn = 3- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“The development of a spelling system in which the same participant with anarthria attempted to silently speak 26 NATO code words — phonetically discriminable words which represent each letter in the alphabet — rather than the 50 common words previously used facilitated access to a larger vocabulary120.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Supportedcites [120] source
The Abstract reports real-time spelling through silent attempts to say 26 NATO alphabet code words, producing sentences with a 1,152-word vocabulary. It also describes the earlier system’s 50-word vocabulary, while the Discussion states that the spelling results expand the previous findings with the same participant. (LLM, conf=0.93)
plaindirectSingle case- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Herff et al.106 further demonstrated that, during continuous speech, phoneme sequences could be decoded provided that synchronized produced acoustics were available to aid model training106.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Supportedcites [106] source
The Introduction describes decoding brain signals during continuously produced speech. Materials and Methods §2.5 says phone labels were derived from acoustic recordings and imposed on the ECoG data; §2.9 describes training the neural phone models. The Results §3.2 reports decoding phones from continuous speech production. (LLM, conf=0.91)
boostedconditional- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“They first built a brain-to-speech lookup library during training by associating 150 ms neural activity segments with the synchronized 150 ms of audible speech.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Partly supportedcites [112] sourceExperimental — accuracy not yet measured
The Materials and Methods sections on Audio Signal Processing and Decoding Approach describe aligned neural features and 150 ms audio units selected from training data. But ECoG Signal Processing specifies 450 ms neural windows, not 150 ms neural segments. Page numbers are unavailable in our copy. (LLM, conf=0.90)
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Multiple studies have demonstrated that isolated phonemes can be decoded from neural activity17,106,107,118 (Fig. 1; Mugler et al. 2014 and Herff et al. 2015).”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Partly supportedcites [17] sourceExperimental — accuracy not yet measured
The ‘Phonetic Organization of Spatial Patterns’ section reports distinct neural activity patterns for consonant articulator groups and for three vowels. However, the Methods describe participants producing consonant–vowel syllables, and the reported analysis clusters neural patterns rather than demonstrating decoding of isolated phonemes. No page number is available in our copy. (LLM, conf=0.83)
- Supportedcites [106] source
The Introduction reports multiple studies that classified isolated phones or phonemes from neural activity, including work by Formisano, Blakely, and Brumberg. This supports the broad claim; the cited paper need not itself be one of the isolated-phoneme studies. (LLM, conf=0.92)
- Partly supportedcites [107] sourceExperimental — accuracy not yet measured
The Introduction reports that earlier studies classified phonemes isolated from words, supporting the general claim. But the Discussion says Mugler et al.’s own study classified individual phonemes within word production, not isolated phonemes, so it does not support citing this study as an example of that specific result. (LLM, conf=0.86)
- Supportedcites [118] source
The Discussion reports that phonemes spoken in clear isolation produced phoneme-specific cortical activity patterns, with 75.5% five-class classification accuracy. It also reports decoding results from several earlier studies, supporting the citing sentence’s reference to multiple studies. (LLM, conf=0.93)
boostedcorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Herff et al.112 developed a concatenative synthesizer to transform neural activity recorded from SEEG into audible speech (Fig. 1; Herff et al. 2019).”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Partly supportedcites [112] sourceExperimental — accuracy not yet measured
The Decoding Approach and Results sections describe a unit-selection method that concatenated speech units to generate audible speech from neural activity. But Materials and Methods (Data Recording) says the activity was recorded with an ECoG electrode grid, not SEEG; the recording modality in the claim is incorrect. The indexed passages provide no page numbers. (LLM, conf=0.92)
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Voice-conversion models require as little as 3 s of recorded speech126; however, larger personalized training datasets can be leveraged if the individual has additional pre-injury audio recordings.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Partly supportedcites [126] sourceExperimental — accuracy not yet measured
Section 5 (Zero-Shot Voice Conversion; page not provided) reports voice-conversion experiments using target-speaker reference recordings longer than 3 seconds. This supports the use of short recordings, but does not establish that exactly 3 seconds is sufficient or that voice-conversion models generally require only that amount. (LLM, conf=0.84)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Card et al.16 trained a personalized TTS model on a small corpus of pre-injury speech, which is an alternative to using a voice-conversion model.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Partly supportedcites [16] sourceExperimental — accuracy not yet measured
The Abstract (page unavailable) says the decoded words were vocalized with text-to-speech software designed to sound like the participant’s pre-ALS voice. Our copy does not say that Card et al. trained that TTS model on a small corpus of pre-injury speech or compare it with voice conversion. The short training times reported in the Abstract concern the neural speech decoder, not the TTS voice. (LLM, conf=0.85)
plainSingle case- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“One line of work involves developing models that learn an underlying manifold (low-dimensional spaces that capture neural population dynamics) structure for a given behaviour103,104,121, which can then be used to stabilize MEA activity over time.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Supportedcites [103] source
The ‘Manifold-based stabilization of neural population activity’ section describes estimating a low-dimensional manifold that captures movement intent, aligning it across time, and passing the stabilized neural activity to a decoder. This supports the claim about using a learned manifold structure to stabilize activity recorded from an electrode array over time. No page number is provided in our copy. (LLM, conf=0.91)
- Partly supportedcites [104] sourceExperimental — accuracy not yet measured
In “Leveraging manifolds and dynamics to stabilize iBCI decoding” (page unavailable), the paper describes learning a neural manifold and updating the mapping from recorded activity to that manifold across sessions. It supports stabilization of decoding despite changes in electrode recordings, but does not say the method stabilizes MEA activity itself. (LLM, conf=0.84)
- Partly supportedcites [121] sourceExperimental — accuracy not yet measured
In “Using latent factors and dynamics to increase BMI performance” (page unavailable), the paper describes models of low-dimensional neural activity that help maintain BMI decoding performance across recording sessions and neuron loss. It does not say these models stabilize the MEA-recorded activity itself; rather, they help decoding remain robust as recordings change. (LLM, conf=0.83)
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“A second approach, specific to text decoding, involves leveraging large language models to automatically recalibrate MEA-based text neuroprostheses.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Partly supportedcites [9] sourceExperimental — accuracy not yet measured
The Daily decoder retraining section reports an offline test in which a language model corrected errors and enabled automatic, unsupervised recalibration of a handwriting decoder. It does not identify that model as a large language model, and it says the approach still needs online confirmation. (LLM, conf=0.88)
- Partly supportedcites [16] sourceExperimental — accuracy not yet measured
The Methods section (‘Decoding speech’; page unavailable) describes an intracortical brain-to-text neuroprosthesis that uses language models to turn predicted phonemes into sentences. It separately says data from different days were combined to calibrate the decoder. The Discussion (page unavailable) says recalibration used prompted sentences with ground-truth labels; the paper does not support the claim that large language models automatically performed recalibration. (LLM, conf=0.86)
- Supportedcites [105] source
The Introduction (page not specified) says CORP uses language models to correct text outputs and automatically retrains the decoder on those corrected outputs. Methods §3.1 identifies GPT2-XL as a large language model used in the text-decoding system and says neural activity was recorded with microelectrode arrays. Together, these passages support the paraphrase. (LLM, conf=0.91)
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Metzger et al.14 first used this voice-conversion approach to personalize synthesized speech for an individual with vocal-tract paralysis using speech samples recorded before the injury.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Supportedcites [14] source
The Speech synthesis section says the authors used a voice-conversion model trained on speech recorded before the participant’s injury to produce her personalized synthetic voice. The Methods specify that the sample came from a pre-injury video, while the Abstract describes her severe vocal paralysis. The Speech synthesis section also says neural speech synthesis had not previously been shown in someone with paralysis, supporting ‘first’ in this context. (LLM, conf=0.90)
plainSingle case- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Furthermore, self-perception of speech is an important component to speech-motor control122,123.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Supportedcites [122] source
The Conclusions state that strong evidence shows sensory feedback is used for online speech control at a cortical level. They also describe auditory and somatosensory feedback processing during speaking, supporting the claim that perceiving one's own speech is important to speech-motor control. (LLM, conf=0.90)
- Supportedcites [123] source
The Conclusion says the CNS actively processes sensory feedback during speaking and uses it to detect deviations from expected speech output. The section on adding feedback control also explains how speakers use auditory and somatosensory feedback to detect and correct changes in their vocal production. Together, these support the paraphrase that perceiving one's own speech is important to speech-motor control. (LLM, conf=0.92)
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“A language model combined the neural-based probabilities of each word with linguistic likelihoods of word sequences, improving WERs of the decoded sentences12 (Fig. 1; Moses et al. 2021; Supplementary Table 1).”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Supportedcites [12] source
The Methods (Modeling) describe a word classifier that derives word probabilities from neural activity and a Viterbi decoder that combines those probabilities with word-sequence probabilities from a natural-language model. The Results (Sentence Decoding) report a lower median word error rate with language modeling (25.6%) than without it (60.5%). (LLM, conf=0.94)
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Recent works have achieved large-vocabulary decoding while maintaining decoding speeds closer to natural speech, by using CTC loss.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Partly supportedcites Metzger et al. 2023 sourceExperimental — accuracy not yet measured
The Overview and Methods > Text decoding describe training the decoders with CTC loss, and Text decoding reports real-time decoding with a 1,024-word vocabulary. However, the 39,378-word result was evaluated offline, while the reported rates of 89.9 and 101 words per minute were for smaller, predefined phrase sets. The passages do not establish that the largest-vocabulary decoding maintained speeds close to natural speech. (LLM, conf=0.82)
- Supportedcites Willett et al. 2023 source
The Decoding attempted speech section describes an RNN with a blank token associated with connectionist temporal classification training, combined with a 125,000-word language model. It reports decoding at 62 words per minute; the Abstract says this begins to approach natural conversation speed (160 words per minute). The paper does not isolate CTC loss as the cause of that performance. (LLM, conf=0.88)
- Partly supportedcites Card et al. 2023 sourceExperimental — accuracy not yet measured
The Results (Online decoding performance; page unavailable) report decoding with a vocabulary of over 125,000 words at the participant’s self-paced rate of 31.6 words per minute. The Methods (Decoding speech; page unavailable) describe phoneme prediction and language models, but the searched passages do not establish use of CTC loss. The Discussion calls the participant’s speaking rate slow, so the paper does not substantiate the comparison to natural speech speed. (LLM, conf=0.85)
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“By using the connectionist temporal classification (CTC) loss119, which scores the loss between predictions from an input sequence (such as neural activity) and an output sequence (such as linguistic subunits) without requiring their alignment, Sun et al.109 trained an RNN to decode character sequences from brain activity during sentence production.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Not checked — no full text of the sourcecites [119] source
- Partly supportedcites [109] sourceExperimental — accuracy not yet measured
The Methods (Brain2Char architecture; page not provided) describe Bi-LSTM recurrent layers in the neural-feature encoder and CTC loss for decoding character sequences from brain recordings. But the text decoder itself uses dilated CNN layers, so describing the trained decoder simply as an RNN misstates that part of the architecture. (LLM, conf=0.86)
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
Every source the paper cites for this claim was checked and supports it.
In the paper
“MEA-based decoders, including text decoders9,15, have traditionally required day-to-day recalibration owing to signal instabilities and drift, requiring additional dedicated time from the user.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Supportedcites [9] source
In “Daily decoder retraining,” the paper says it followed standard practice by retraining its handwriting text decoder each day using calibration data. It explains that retraining accounts for changes in neural recordings over time and that collecting calibration data burdens the user. (LLM, conf=0.93)
- Supportedcites [15] source
In “Decoding attempted speech,” the paper describes daily training to account for across-day changes in neural activity, taking dedicated participant time. The Discussion says that adapting to changes across several days without requiring the user to pause and recalibrate remains future work. Its offline finding of reasonable performance without new-day training qualifies the need for recalibration, but does not negate the claim about the practice used in this demonstration. (LLM, conf=0.88)
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Their approach further improved speech-synthesis performance by leveraging the underlying articulatory organization of the SMC.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Supportedcites [110] source
The Speech decoder design section says the decoder uses an intermediate articulatory representation because ventral sensorimotor cortex activity predominantly encodes articulatory kinematics. The Decoder characteristics section reports a clear performance advantage for this approach over decoding acoustics directly from the same neural signals. Page numbers were not available in our copy. (LLM, conf=0.93)
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“This approach has primarily been applied in the context of upper limb decoders103 but may scale to speech neuroprostheses in the future.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Supportedcites [103] source
The Methods section reports that the monkeys’ arrays were implanted in the proximal arm region of primary motor cortex, and the paper describes testing its manifold-based stabilizer with a movement-decoding cursor BCI. This supports the citation as an example of the approach in an upper-limb-related decoder; the sentence does not attribute its speculation about speech neuroprostheses to this source. (LLM, conf=0.85)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“the latter of which may be increased without substantially limiting performance14,15.”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Evaluation and standardization
What the sources say
- Supportedcites [14] sourceExperimental — accuracy not yet measured
The Text decoding section reports that decoding remained robust as the vocabulary grew from 1,024 to 39,378 words. The larger-vocabulary results were obtained through offline simulations, as the Methods clarify; within that tested setting, the source supports the claim. (LLM, conf=0.91)
- Partly supportedcites [15] sourceExperimental — accuracy not yet measured
In “Design considerations for speech BCIs” (page unavailable), word error rates levelled off at about 1,000 words, supporting the idea that further vocabulary growth may have little effect. But the paper also says only very small vocabularies retained a large accuracy advantage: in “Decoding attempted speech” (page unavailable), the error rate rose from 9.1% with 50 words to 23.8% with 125,000 words. The unqualified claim understates that loss of performance. (LLM, conf=0.86)
hedgedconditional- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Previous work has demonstrated that first decoding articulatory features and transforming them to acoustic features improved the quality of synthesized speech (compared with decoding acoustics directly from neural data)110.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Supportedcites [110] sourceExperimental — accuracy not yet measured
The Neural decoder methods describe a two-stage model that decodes articulatory features from ECoG signals before transforming them into acoustic features. Decoder characteristics reports better performance for this model than for a decoder that maps the same signals directly to acoustics; the supplied passages do not specify page numbers. (LLM, conf=0.93)
boosteddirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“To resolve this, proxy waveforms, such as waveforms generated from TTS or recorded by able speakers and used during model training, can be used as references in MCD computations14.”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Evaluation and standardization
What the sources say
- Supportedcites [14] source
In Speech synthesis and Methods > Speech synthesis (page not available), the paper says its reference waveforms came from a recruited speaker for the AAC sets and a text-to-speech model for the 1024-word set. It used those waveforms to create reference speech-unit sequences for training, and the Evaluation subsection says MCD was computed between decoded and reference waveforms. (LLM, conf=0.91)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“ASR can also be used to transcribe synthesized speech into text to compute WERs; however, ASR systems are typically trained to recognize natural waveforms that are not artificially synthesized153.”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Evaluation and standardization
What the sources say
- Partly supportedcites [153] sourceExperimental — accuracy not yet measured
The Methods (Automated speech recognition models; page unavailable) say that the Wav2vec 2.0 model was trained on speech corpora, and the paper applies ASR to BCI-synthesized speech. Our copy does not establish that ASR systems generally are trained on natural waveforms rather than artificially synthesized ones. (LLM, conf=0.78)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“By contrast, in previous multiparticipant studies in able speakers, participants had near-identical speech capacity106,108,136–138.”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Evaluation and standardization
What the sources say
- Partly supportedcites [106] sourceExperimental — accuracy not yet measured
The Subjects and Experiment sections (2.1 and 2.3; page unavailable) describe seven participants who produced speech by reading text aloud, supporting the characterization of this as a multiparticipant study of speakers. They do not assess or establish that the participants had near-identical speech capacity. (LLM, conf=0.83)
- Partly supportedcites [108] sourceExperimental — accuracy not yet measured
The Methods (Task; page not provided) say that participants read sentences aloud, supporting the description of this as a multiparticipant study of able speakers. Neither that section nor the Methods participant description (page not provided) establishes that their speech capacities were near-identical. (LLM, conf=0.82)
- Not checked — no full text of the sourcecites [136] source
- Partly supportedcites [137] sourceExperimental — accuracy not yet measured
The Results describe seven participants completing overt word-production tasks, and report that they could produce both real words and pseudowords. Neither the Results nor Materials and Methods compare participants’ speech capacities or establish that they were near-identical; a search for speech ability and deficits found no such comparison. (LLM, conf=0.78)
- Partly supportedcites [138] sourceExperimental — accuracy not yet measured
The Online Methods (Participants; page unavailable) describe 16 participants selected for cognitive and language abilities in the average range or above, supporting the description of a multiparticipant study of able speakers. That criterion does not establish that their speech capacities were near-identical. (LLM, conf=0.82)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Studies with both ECoG14 and MEAs15,16,52 in individuals with paralysis have demonstrated the feasibility of decoding articulatory features”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Supportedcites [14] source
This source supports the ECoG portion of the sentence: the Overview of multimodal speech-decoding system describes a participant with severe paralysis and anarthria whose ECoG signals were used to predict articulator movements. The section Articulatory representations drive decoding also reports articulatory-feature encoding in those signals. The separate MEA citations are not attributed to this source. (LLM, conf=0.93)
- Supportedcites [15] source
The Main section describes a participant with ALS who could no longer produce intelligible speech and records from intracortical microelectrode arrays. It reports decoding orofacial movements and phonemes from those recordings, supporting the MEA portion of the citing sentence; the ECoG portion is attached to a different citation. (LLM, conf=0.91)
- Partly supportedcites [16] sourceExperimental — accuracy not yet measured
The Methods (Decoding speech; page unavailable) describe using cortical neural activity to predict attempted phonemes, and the Methods (Recording array implant locations and decoding contributions; page unavailable) identify four microelectrode arrays implanted in speech-related cortex. The study supports the feasibility of MEA-based speech decoding in a person with paralysis, but it reports decoding phonemes rather than articulatory features specifically. (LLM, conf=0.83)
- Partly supportedcites [52] sourceExperimental — accuracy not yet measured
The Results (‘Speech can be decoded from intracortical activity on individual trials’; page unavailable) report decoding spoken syllables and words from intracortical array recordings in two people with tetraplegia. The paper relates the neural patterns to articulation, but it does not report decoding articulatory features themselves; its Discussion says linking activity to continuous articulatory movements requires future work. This source supports the MEA portion of the sentence, not an independent ECoG demonstration. (LLM, conf=0.84)
boostedcorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources that could be checked couldn’t confirm or dispute this claim.
In the paper
“The performance of TTS systems is often described using mel-cepstral distortion (MCD) and open-ended human-transcribed WER149,150.”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Evaluation and standardization
What the sources say
- Could not verifycites [149] sourceExperimental — accuracy not yet measured
The Approach and Results sections evaluate AAC word prediction using communication rate, keystrokes, and related measures. Searches for mel-cepstral distortion, TTS, and human-transcribed word error rate found no passage addressing how TTS performance is described. (LLM, conf=0.78)
- Not checked — no full text of the sourcecites [150] source
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Could not verify
The cited sources that could be checked couldn’t confirm or dispute this claim.
In the paper
“Delayed auditory feedback disrupts speech production, so an ideal streaming neural-speech synthesizer would output sound with negligible latencies similar to the self-perception of able speech (within a few milliseconds).”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Evaluation and standardization
What the sources say
- Could not verifycites [154] sourceExperimental — accuracy not yet measured
Section 3.8 (p. 9) describes Whisper’s 30-second input chunks and buffered transcription, not the effects of delayed auditory feedback on speech production. Searches for delayed auditory feedback and millisecond-scale synthesis latency found no passage in our copy addressing the attributed claim. (LLM, conf=0.76)
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Could not verify
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Metzger et al. demonstrated that it is possible to use these decoded articulatory features to animate a digital avatar face, which was personalized to resemble the likeness of the individual14 (Fig. 3d).”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Partly supportedcites [14] sourceExperimental — accuracy not yet measured
The Facial-avatar decoding section says decoded articulatory gestures were used to animate the avatar face, and the Discussion calls it a personalized avatar. However, our copy does not say the avatar’s face was personalized to resemble the participant; the explicit resemblance claim in Methods > Speech synthesis concerns her synthetic voice. (LLM, conf=0.84)
boosteddirectSingle case- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The sources with a verdict support this claim; the rest weren’t checked yet, have no full text we could read, or their verdicts aren’t rated yet.
In the paper
“This would likely provide substantial improvements over AAC, which typically do not exceed 15 WPM (refs. 146–148).”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Evaluation and standardization
What the sources say
- Supportedcites [146] source
The Introduction says AAC communication rate is often less than 10 words per minute. The Approach describes 8–10 words per minute as a rate typical of AAC users, supporting the claim that AAC typically does not exceed 15 WPM. (LLM, conf=0.93)
- Not matchedcites [147]
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Not checked — no full text of the sourcecites [148] source
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Generative pretrained transformers145 are a class of language models that are particularly powerful for this application.”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Evaluation and standardization
What the sources say
- Partly supportedcites [145] sourceExperimental — accuracy not yet measured
The Abstract (p. 1) identifies GPT-4 as a Transformer-based model pre-trained to predict the next token, supporting the description of GPTs as language models. The report does not establish that they are particularly powerful for rescoring decoded sequences; searches for rescoring and speech-recognition decoding found no supporting passage. (LLM, conf=0.84)
boosted- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Given that most early studies in individuals with vocal-tract paralysis involve only single participants12–16,64,95,120, great care is needed when comparing results, as the broad spectrum of disease presentations is not sampled”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Evaluation and standardization
What the sources say
- Supportedcites [12] source
The Methods (Participant) describe one man with anarthria after a pontine stroke. The Methods (Study Overview) state that, at the time of writing, the device had been implanted only in that participant. This paper therefore supports its use as an example of a single-participant study; the citation is one of several offered for the broader statement about most early studies. (LLM, conf=0.91)
- Supportedcites [13] source
The Materials and Methods ‘Participant’ section describes one participant with ALS whose speech was mostly unintelligible. The Discussion identifies this as a demonstration involving a participant with impaired articulation and phonation. As one citation in a group of early studies, this paper supports its role as a single-participant example; it does not independently establish the word ‘most.’ (LLM, conf=0.90)
- Supportedcites [14] source
The citation serves as an example of a single-participant study, not as the sole evidence that most early studies had that design. The Discussion explicitly says the results are from only one participant; the supplied copy gives no page number. (LLM, conf=0.90)
- Supportedcites [15] source
The Main section describes one study participant, T12, who has bulbar-onset ALS and cannot produce intelligible speech. The discussion says the decoding results need confirmation in additional participants. This paper therefore supports the citation as an example of a single-participant study, though it does not by itself establish that most early studies used one participant. (LLM, conf=0.90)
- Supportedcites [16] source
The Limitations section says this study involved a single participant, as did other recent clinical-trial reports in implanted speech brain-computer interfaces. It also calls for additional participants to establish how performance varies across individuals, supporting the citing sentence’s point about limited sampling. (LLM, conf=0.87)
- Partly supportedcites [64] sourceExperimental — accuracy not yet measured
The Introduction and Experimental model and subject details describe a study of one participant. But the Discussion calls him “speech-abled,” so this paper does not provide an example of a study in an individual with vocal-tract paralysis. It also does not establish that most early studies used single participants. (LLM, conf=0.90)
- Partly supportedcites [95] sourceExperimental — accuracy not yet measured
The Experimental Section (Clinical Trial and Participant; page unavailable) reports data from one participant, so this paper supports the single-participant aspect of the claim. It describes severe dysarthria from ALS but says the participant could still phonate and articulate; it does not establish vocal-tract paralysis or the broader assertion that most early studies used single participants. (LLM, conf=0.84)
- Supportedcites [120] source
The Participant section describes a person with anarthria who cannot produce intelligible speech. The Reproducibility of experiments section explicitly calls this a pilot study with a single participant, making it an example of the single-participant studies cited in the sentence. (LLM, conf=0.91)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“In addition, articulatory features may be useful as a standalone output space for both speech and non-speech orofacial gestures14,52,113,114 (Fig. 3d).”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Supportedcites [14] source
The Facial-avatar decoding section reports direct decoding of articulatory gestures from neural activity for speech and classification of attempted non-speech orofacial movements. The Articulatory representations drive decoding section also describes articulatory encoding as general-purpose rather than speech-specific. Together, these findings support the hedged suggestion that articulatory features may be useful as an output space for both. (LLM, conf=0.84)
- Partly supportedcites [52] sourceExperimental — accuracy not yet measured
The Results (Speech-related activity in dorsal motor cortex; page unavailable) show that neural activity grouped phonemes by place of articulation and that many electrodes active during speech were also active during non-speech orofacial movements. This supports the relevance of articulatory features to both behaviors, but the paper does not evaluate them as a standalone output space; the Methods say precise movement kinematics were not collected. (LLM, conf=0.81)
- Partly supportedcites [113] sourceExperimental — accuracy not yet measured
The Discussion (page unavailable) reports that distinct articulator and tongue movements can be classified and suggests their neural activity could provide additional BCI control signals. It also says the findings may benefit speech-BCI development, but the study tested isolated movements rather than a standalone articulatory output space for both speech and non-speech gestures. (LLM, conf=0.82)
- Partly supportedcites [114] sourceExperimental — accuracy not yet measured
The Discussion (page not provided) reports that neural activity can distinguish posed facial expressions involving the lips, jaw, nose, and eyebrows. The Conclusion proposes adding those expressions to communication BCIs, supporting a potential output for non-speech gestures. The paper does not establish articulatory features as a standalone output space for speech. (LLM, conf=0.86)
hedgedconditional- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“an additional safeguard is to use a speech-detection algorithm to gate any decoding process, which is highly successful in correctly identifying volitional attempts of users to overtly and silently speak12,120.”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Practical implementation of a speech neuroprosthesis
What the sources say
- Partly supportedcites [12] sourceExperimental — accuracy not yet measured
The Results, ‘Word Detection and Classification’ (page not provided), report that the detector identified 98% of one participant’s attempted word productions, supporting high detection of attempted speech. The study involved a person with anarthria; it does not establish comparable performance for both overt and silent speech across users. It also reports 968 spurious detections, so the result does not establish that detected events consistently represent volitional attempts. (LLM, conf=0.86)
- Partly supportedcites [120] sourceExperimental — accuracy not yet measured
The Speech detection section (page unavailable) describes a real-time model that detects the participant’s silent-speech attempts to engage the spelling system, supporting the gating part of the claim. It does not establish that detection was highly accurate for both overt and silent attempts; the overt-versus-silent results concern classification of attempted code words, not detection of whether speech was attempted. (LLM, conf=0.84)
boosteddirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“For example, higher-density cortical sampling — via an increased number of surface or penetrating electrodes — has been shown to increase the performance of decoding models14–16,175,176.”
§ The speech neuroprosthesis > Future directions
What the sources say
- Supportedcites [14] source
In “Articulatory representations drive decoding,” the paper reports degraded decoding performance in an offline simulation of lower-density electrode sampling and identifies this as a benefit of high-density cortical recording. Its evidence concerns a surface ECoG array; the sentence’s other citations may address penetrating electrodes. (LLM, conf=0.90)
- Partly supportedcites [15] sourceExperimental — accuracy not yet measured
In “Design considerations for speech BCIs” (Fig. 4b; page unavailable), the paper reports that offline decoding accuracy improved as more intracortical electrodes were included. It suggests denser arrays may improve future accuracy, but does not demonstrate the claim for surface electrodes or test a newly increased physical electrode density. (LLM, conf=0.84)
- Partly supportedcites [16] sourceExperimental — accuracy not yet measured
The Methods say this study doubled the number of implanted microelectrodes relative to an earlier study, and the Discussion reports higher decoding performance. But the Discussion calls electrode count only a possible contributing factor and names other changes that may have helped; it does not establish that more electrodes caused the improvement. This paper also does not test increased surface-electrode sampling. (LLM, conf=0.86)
- Not checked — no full text of the sourcecites [175] source
- Supportedcites [176] source
The Results section reports that decoding performance increased as electrode density increased on the human superior temporal gyrus. The Conclusions likewise state that higher-density surface recordings enhanced neural decoding within individual patients. This paper supports the surface-electrode example; it does not study penetrating electrodes. (LLM, conf=0.91)
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“A similar tradeoff can be facilitated by having a spelling framework available on the neuroprosthesis, either using attempted speech120 or handwriting9, which would allow users to spell out arbitrary sentences with high accuracy (albeit with substantially lower speeds).”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Practical implementation of a speech neuroprosthesis
What the sources say
- Supportedcites [120] source
The Task design and Overview of the real-time spelling pipeline sections describe an attempted-speech system used to spell arbitrary, unprompted sentences. The Generalizability to larger vocabularies and alternative tasks section reports a median character error rate of 14.8% for conversational messages, and the Discussion describes generalizable spelling as slower than rapid full-word decoding. This source supports the attempted-speech part of the sentence; handwriting has a separate citation. (LLM, conf=0.88)
- Partly supportedcites [9] sourceExperimental — accuracy not yet measured
The Discussion (page unavailable) says the handwriting BCI lets its user express any sentence, and Decoding handwritten sentences reports high accuracy. The paper does not establish the claimed tradeoff of substantially lower speed; it describes 90 characters per minute as a new BCI speed benchmark. The attempted-speech alternative has a separate citation and is not assessed here. (LLM, conf=0.83)
hedgeddirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“As long and recurring training periods for speech neuroprostheses might discourage long-term use158,159, it is essential to develop decoders that are stable over time or can be quickly recalibrated102,104,105 to avoid dedicated and high-effort re-training periods inconveniencing users.”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Practical implementation of a speech neuroprosthesis
What the sources say
- Not checked — no full text of the sourcecites [158] source
- Partly supportedcites [159] sourceExperimental — accuracy not yet measured
The Results (page not available) report users’ concerns that BCI setup was time-consuming and training-intensive for caregivers, particularly with frequent staff turnover. The paper does not say that speech-neuroprosthesis users face long, recurring training periods or that such periods discourage long-term use. (LLM, conf=0.78)
hedgedconditional- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
In the paper
“During online inference, the decoded discrete acoustic-speech unit sequences were vocoded into sentence-level speech that was intelligible to untrained human listeners (Supplementary Table 1).”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Not matchedcites Supplementary Table 1
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Not matched
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“These persisting neural representations provide a substrate for decoding intended articulation and vocalization, even if these motor commands no longer reach their destination in the vocal tract.”
§ The speech neuroprosthesis > Neurological speech disorders impairing communication
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption: In individuals with dysarthria or anarthria, cerebral cortex neuronal populations that contain representations of articulatory movements may remain intact.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Although restoring natural speech is a long-term goal, speech neuroprostheses already have performance levels that surpass communication rates offered by current assistive-communication technology.”
§ The speech neuroprosthesis > Abstract
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: unclear — It is unclear why the paper says it holds.
In the paper
“However, all invasive recording techniques share risks related to the anaesthesia, craniotomy and dural opening in the implantation surgery and the possibility of infection81.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Neural-recording interfaces
What the sources say
- Not matchedcites [81]
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Not matched
In the paper
“ECoG has been used clinically for decades for localizing seizures and for brain mapping in surgical work-up for epilepsy treatment90.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Neural-recording interfaces
What the sources say
- Not checked — no full text of the sourcecites [90] source
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Thus, the stable acquisition, broad coverage and high performance of ECoG make it a strong candidate for a clinically viable neuroprosthesis.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Neural-recording interfaces
boosteddirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Together, these studies all highlighted a large region of cortex, primarily on the SMC, that has been subsequently targeted in the first clinical trials of speech neuroprostheses.”
§ The speech neuroprosthesis > Cortical encoding of speech features
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Thus, speech and language features encoded in the STG, SMG and Broca’s area could potentially be leveraged for speech decoding in disorders in which articulatory representations on the SMC are not intact (Box 1).”
§ The speech neuroprosthesis > Cortical encoding of speech features
hedgedconditional- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
The paper states this without a source or data behind it here, so there is nothing to check it against. The quote shows it is in the paper.
In the paper
“The previously described speech-synthesis approaches required precise alignment between neural activity and ground-truth acoustics.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Despite their successes, these approaches are all limited by restrictive, predefined vocabulary sizes.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“A key advantage of using CTC loss is that the individual can more naturally produce a phrase in this approach, without having to associate isolated speech attempts with individual words.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Overall, this has led to state-of-the-art decoding performance for individuals with paralysis.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Because speech is variable across people and expressive, neuroprostheses that synthesize speech are well suited for personalization (Fig. 3c).”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
None of the sources cited for this claim could be matched to a paper in Crossref, so there is nothing to check it against.
In the paper
“Language models then mapped decoded phoneme sequences into words and sentences (Fig. 3b and Supplementary Table 1).”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Not matchedcites Fig. 3b
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Not matchedcites Supplementary Table 1
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Not matched
The paper states this without a source or data behind it here, so there is nothing to check it against. The quote shows it is in the paper.
In the paper
“A voice-conversion model can be applied to convert decoded speech waveforms into personalized waveforms that resemble the likeness of the user.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Furthermore, decoders can generalize to larger vocabularies by targeting low-dimensional subword units (such as phonemes) that can be built into a much higher-dimensional space of words and sentences.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Thus, promising avenues exist to create MEA-based text neuroprostheses that do not require daily recalibration time from the user.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“However, individuals with severe cases of dysarthria or anarthria may lack the ability to make any intelligible vocalizations, especially for longer sentences.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
hedgedconditional- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Similar to text decoding, a way to circumvent the need for alignment is to use CTC loss to train models that map input sequences of neural activity to output sequences of acoustic signals that may then be vocoded into speech.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
hedgeddirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“However, as described earlier, a major challenge of creating speech neuroprostheses for individuals with vocal-tract paralysis is the lack of intelligible ground-truth acoustic signals to align with the neural data during training.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“This allows the RNN to be automatically recalibrated without additional training labels or dedicated time on the part of an individual with paralysis.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“However, for individuals who cannot speak, there may be no way to reliably align neural activity with ground-truth produced acoustics.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
hedged- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A claim is checked only against the cited paper’s full text, never its abstract, and none is available for its sources.
In the paper
“Given the strong link between speech acoustics and underlying vocal-tract movements, articulation constitutes another relevant feature space (Fig. 3d) for speech decoding and speech synthesis128,129.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
boostedcorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
In the paper
“When attempting to speak, any audible vocalizations that an individual with anarthria produces are generally unintelligible140 owing to their lack of coordinated vocal-tract control (Fig. 4a).”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Evaluation and standardization
What the sources say
- Not matchedcites [140]
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Not matched
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“However, a strong indicator that a speech neuroprosthesis would allow the user to produce words beyond the training vocabulary is high performance on an evaluation set that is reflective of natural language, containing words not seen by the speech-decoding model during training.”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Evaluation and standardization
boostedcorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
In the paper
“It is preferred to correlation-based metrics, such as computing a Pearson correlation between ground truth and decoded acoustic waveforms, as MCD weighs the features in the produced sound that are most perceptually relevant for conveying information151.”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Evaluation and standardization
What the sources say
- Not checked — no full text of the sourcecites [151] source
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
The paper states this without a source or data behind it here, so there is nothing to check it against. The quote shows it is in the paper.
In the paper
“Most prominently, language models can be used to rescore sequences of linguistic features (such as phonemes or words) that are decoded from neural activity. Language models perform this re-scoring by evaluating how likely decoded sequences are to occur in natural language.”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Evaluation and standardization
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The use of the MCD metric is, however, complicated by the lack of precisely aligned ground-truth speech when working with individuals who cannot speak.”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Evaluation and standardization
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
In the paper
“Although current speech neuroprostheses do not decode or target internal monologue (Fig. 4a), an additional safeguard is to use a speech-detection algorithm to gate any decoding process, which is highly successful in correctly identifying volitional attempts of users to overtly and silently speak12,120.”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Practical implementation of a speech neuroprosthesis
What the sources say
- Not matchedcites Fig. 4a
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Not matched
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Overall, a positive correlation exists between decoder performance and electrode density and/or the relevant coverage afforded by the corresponding neural interface, emphasizing that improvements to these two hardware aspects can lead to improved speech brain computer interfaces, regardless of the specific recording modality.”
§ The speech neuroprosthesis > Future directions
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Overall, a positive correlation exists between decoder performance and electrode density and/or the relevant coverage afforded by the corresponding neural interface, emphasizing that improvements to these two hardware aspects can lead to improved speech brain computer interfaces, regardless of the specific recording modality.”
§ The speech neuroprosthesis > Future directions
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Probably not a substantial claim. 35
- It may state two separate claims as one. 50
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Another important acoustic feature is the fundamental frequency at which the laryngeal vocal folds resonate during speech, which forms the basis of pitch and can be estimated from the acoustic waveform, using algorithms that infer the periodicity of a temporal signal28,29 (Fig. 2e).”
§ The speech neuroprosthesis > Overview of speech features
What the sources say
- Not checked — no full text of the sourcecites [28] source
- Partly supportedcites [29] sourceExperimental — accuracy not yet measured
The Introduction section “Larynx and its peripheral nervous control” says vibrating vocal folds produce the basic sound wave and that changes in vocal-fold length modulate fundamental frequency. The paper also discusses pitch modulation, but the passages do not explicitly connect fundamental frequency to perceived pitch or describe estimating it from an acoustic waveform using periodicity algorithms. No page number is available in our copy. (LLM, conf=0.86)
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“However, communication rates for neuroprostheses controlled by attempted hand movements remain slower and less expressive than natural speech10.”
§ The speech neuroprosthesis > Introduction
What the sources say
- Partly supportedcites [10] sourceExperimental — accuracy not yet measured
In “Toward a Speech Neuroprosthesis” (page unavailable), the paper says neuroprostheses that decode arm-movement signals to control a spelling cursor achieve under 10 words per minute, versus 120–150 words per minute for natural speech. It describes speech features such as prosodic intonation, but does not directly establish that hand-movement-controlled neuroprostheses are less expressive. (LLM, conf=0.85)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Speech enables efficient and rapid communication at rates of around 150–200 words per minute (WPM) in conversational settings10.”
§ The speech neuroprosthesis > Neurological speech disorders impairing communication
What the sources say
- Partly supportedcites [10] sourceExperimental — accuracy not yet measured
The 'Toward a Speech Neuroprosthesis' section describes speech as efficient, rapid communication and gives natural speaking rates of 120–150 words per minute in healthy speakers. It does not support the cited range of 150–200 words per minute or specify that rate for conversational settings. No page number is supplied in our copy. (LLM, conf=0.88)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Furthermore, the midPrCG is involved across many speech functions and does not appear strongly tuned to — that is, consistently and exclusively activated by — a single articulatory group or movement (Fig. 2a), potentially instead having a critical role in speech-motor planning47–50.”
§ The speech neuroprosthesis > Cortical encoding of speech features
What the sources say
- Not checked — no full text of the sourcecites [47] source
- Not checked — no full text of the sourcecites [48] source
- Not checked — no full text of the sourcecites [49] source
- Partly supportedcites [50] sourceExperimental — accuracy not yet measured
The Discussion (Lessons; page unavailable) proposes that the middle precentral gyrus helps sequence motor plans independently of the articulators involved, and the Observations describe effects across speaking, reading, and writing. The case does not assess whether the region is consistently and exclusively activated by a particular articulatory group or movement, so it cannot establish the claim about tuning. (LLM, conf=0.84)
hedgedconditional- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Subsequent work has expanded this initial demonstration by directly decoding cortical activity into audible speech13,14 and allowing more generalizable and rapid text decoding14–16 in individuals with vocal-tract paralysis (Fig. 1).”
§ The speech neuroprosthesis > Introduction
What the sources say
- Supportedcites [13] source
The Discussion reports that a participant with ALS and impaired articulation and phonation produced acoustic speech through online decoding of cortical ECoG activity. This supports the audible-speech part of the sentence for which this paper is cited; the separate text-decoding claim has additional citations and is not attributed to this paper alone. (LLM, conf=0.88)
- Supportedcites [14] source
The Abstract reports real-time decoding of cortical recordings into synthesized speech audio and rapid, large-vocabulary text in a participant with severe vocal paralysis. The Discussion says text decoding was faster than previous BCIs and used a vocabulary 20 times larger than the authors’ previous system. Page numbers are not provided in our copy. (LLM, conf=0.92)
- Supportedcites [15] source
The text-decoding portion of the citing sentence is supported. In “Decoding attempted speech,” the paper reports real-time decoding with a 125,000-word vocabulary suitable for general English; the Discussion reports a speed of 62 words per minute. The study participant had ALS and could no longer speak intelligibly. Page numbers are unavailable in our copy. (LLM, conf=0.89)
- Partly supportedcites [16] sourceExperimental — accuracy not yet measured
The Abstract reports rapid cortical-activity-to-text decoding with a 125,000-word vocabulary, and Results > Online decoding performance says the decoder generalized to new words. But Limitations > Decoding speech in patients with other neurological impairments says this participant had ALS-related dysarthria, retained some voluntary bulbar muscle control, and that performance in people with anarthria is unknown. The cited work therefore supports the text-decoding advance, not its application to individuals with complete vocal-tract paralysis. (LLM, conf=0.86)
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Dichter et al.44 defined dorsal and ventral laryngeal cortical representations, with the dorsal one controlling vocal pitch for prosodic intonation in English44 and, interestingly, lexical tone production in Mandarin46.”
§ The speech neuroprosthesis > Cortical encoding of speech features
What the sources say
- Supportedcites [44] source
The Results sections on natural speech intonation and dLMC/vLMC pitch components describe an English word-emphasis task, identify dorsal and ventral laryngeal motor cortex representations, and link dLMC activity to pitch accents. The Mandarin lexical-tone statement is attached to citation 46, not to Dichter et al. (citation 44), so it is outside this verdict. (LLM, conf=0.92)
- Supportedcites [46] source
The Results section, “Intraoperative direct electrical stimulation of dLMC elicits pitch rising and lowering” (page not provided), reports that stimulating dorsal laryngeal motor cortex changed pitch during Mandarin tone production. The Introduction (page not provided) also describes dorsal and ventral laryngeal motor regions and cites earlier work on dorsal pitch control in English; the citing sentence assigns that earlier work to citation 44, not this paper. (LLM, conf=0.92)
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Neural activity in the STG encodes acoustic features — for example, recurring speech sound patterns in the mel-spectrogram — during speech perception53,54 and may be used to reconstruct perceived speech55,56.”
§ The speech neuroprosthesis > Cortical encoding of speech features
What the sources say
- Supportedcites [53] source
The Introduction reports that speech-selective STG regions are tuned to spectrotemporal acoustic patterns and that speech signals can be reconstructed from STG activity using STM-based models. The reconstruction result is reported from an earlier study, not presented as a new result of this paper. (LLM, conf=0.88)
- Supportedcites [54] source
The paper reports STG responses to speech sounds and finds that phonetic-feature selectivity reflects tuning to acoustic spectrotemporal cues (section labeled Abstract; page not provided). In the citing sentence, this source supports the encoding claim; separate citations 55 and 56 are attached to the reconstruction claim. (LLM, conf=0.91)
- Supportedcites [55] source
The citation is attached to the claim that neural activity may be used to reconstruct perceived speech. The Introduction’s Figure 1 description says listeners heard speech and their evoked auditory-cortex activity was used to reconstruct it; the Results (Neural recordings) confirm that some recording electrodes covered the STG. The mel-spectrogram example is attached to different citations and is not assessed here. (LLM, conf=0.91)
- Supportedcites [56] source
The Discussion reports that key features of continuous speech were accurately reconstructed from STG neural responses using spectrogram-based models. The Results describe participants listening to words while electrodes over the temporal gyrus recorded signals used to reconstruct speech spectrograms. The claim does not imply that the reconstructed speech was intelligible to human listeners. (LLM, conf=0.94)
hedgedconditional- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Although these non-invasive techniques have facilitated decoding of limited vocabularies of imagined words, syllables and speech sounds75,76 and reconstruction of semantic content of speech68 (Fig. 1; Tang et al. 2023), they are not portable — with the exception of EEG — and low spatial or temporal resolution has hindered scaling decoding to larger vocabularies and longer speech segments (senten”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Neural-recording interfaces
What the sources say
- Supportedcites [75] source
Section 2.4 (Choice of Prompts; page unavailable) describes EEG decoding studies using small sets of imagined syllables, vowels and words. Section 4.4 discusses their classification performance. The semantic-content reconstruction clause is attached to a different citation, so it is not attributed to this review. (LLM, conf=0.91)
- Supportedcites [76] source
The Discussion (section 8) reports EEG-based classification of limited sets of imagined words, syllables, vowels, and phonemes. The sentence cites other sources for reconstruction of semantic content, so that is not attributed to this paper. (LLM, conf=0.91)
- Supportedcites [68] source
The Tang et al. citation is attached to the reconstruction of semantic content, not the preceding claim about limited vocabularies. The Introduction describes a decoder using non-invasive fMRI to reconstruct perceived or imagined stimuli in continuous natural language, and the Discussion states that it reconstructs meaning, though often not the exact words. The supplied passages have section labels but no page numbers. (LLM, conf=0.94)
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“The first efforts towards a speech neuroprosthesis for an individual with vocal-tract paralysis, developed by Phil Kennedy and Frank Guenther, used an intracortical neurotrophic microelectrode placed in the precentral gyrus, which allowed above-chance synthesis of vowel formants84 and decoding of phonemes85 (Fig. 1; Guenther et al. 2009 and Brumberg et al. 2011).”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Neural-recording interfaces
What the sources say
- Supportedcites [84] source
Split into 2 atoms — 2 confirmed, 0 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.
- Supportedcites [85] source
Split into 2 atoms — 2 confirmed, 0 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.
plaindirectSingle case- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Previous demonstrations of intracortical insertion of MEAs showed injury and neuronal loss at the implantation location99, as well as signal instabilities contributing to signal loss and performance decline over time100,101.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Neural-recording interfaces
What the sources say
- Supportedcites [99] source
The Discussion reports a 63% decrease in neuron density around implanted Utah-array shanks compared with nearby non-implanted tissue in a non-human primate, describing it as a drastic disruption of neural populations. Citation 99 is attached to the injury and neuronal-loss clause; the separate claim about signal instabilities has other citations. (LLM, conf=0.90)
- Supportedcites [100] source
The citation to this paper is attached to the signal-instability and declining-performance portion of the sentence, not the injury and neuronal-loss portion. Section 4.6 reports sporadic recordings and progressive decay in signal quality, while sections 5.3–5.4 discuss causes of signal loss and instability. The copy provides no page numbers for these passages. (LLM, conf=0.87)
- Supportedcites [101] source
Citation 99 is attached to the injury and neuronal-loss clause, not the later signal-instability clause. The Introduction reports tissue damage near an implanted human array and describes neuronal loss around electrode tips from device insertion. (LLM, conf=0.88)
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“The high-γ band of the local field potential (70–150 Hz) correlates with neuronal firing rates underneath each ECoG electrode86,87 and captures cortical representations of acoustic and articulatory features relevant to speech17,88,89 (Fig. 1; Crone et al. 1998, 2001).”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Neural-recording interfaces
What the sources say
- Partly supportedcites [86] sourceExperimental — accuracy not yet measured
The Discussion (page not provided) reports that high-gamma LFP activity correlates with neuronal firing near a microelectrode in macaque visual cortex. It does not establish that relationship underneath each ECoG electrode, and the Introduction defines high gamma as 80–200 Hz rather than 70–150 Hz. The speech-feature clause has separate citations in the citing sentence and is not attributed to this source. (LLM, conf=0.88)
- Partly supportedcites [87] sourceExperimental — accuracy not yet measured
The Discussion reports that LFP high-gamma power is tightly correlated with firing rate near a microelectrode. It distinguishes this measured result from its ECoG model, in which high-gamma power is more sensitive to firing synchrony than to firing rate; it does not establish the claimed correlation underneath each ECoG electrode. The speech-feature clause has separate citations and is not attributed to this source. (LLM, conf=0.87)
- Partly supportedcites [17] sourceExperimental — accuracy not yet measured
The paper’s ‘Phonetic Organization of Spatial Patterns’ section links cortical activity to articulators and vowel states, supporting the articulatory part of the claim. Its ‘vSMC Physiology During Syllable Production’ section also says high-gamma activity correlates with multi-unit firing rates, but specifies 85–175 Hz rather than 70–150 Hz. The cited passages do not establish that this paper found cortical representations of acoustic features; its acoustic analysis describes the recorded speech sounds. (LLM, conf=0.86)
- Not checked — no full text of the sourcecites [88] source
- Not checked — no full text of the sourcecites [89] source
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Functional MRI, by contrast, has higher spatial specificity but lower temporal resolution than EEG and MEG as it measures changes in cerebral blood flow as a correlate of neural activity74.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Neural-recording interfaces
What the sources say
- Partly supportedcites [74] sourceExperimental — accuracy not yet measured
The Introduction (page not provided) says BOLD fMRI can localize activated brain areas and that its response lags a stimulus by 1–2 seconds. It does not compare fMRI’s spatial or temporal resolution with EEG or MEG. It describes changes in blood oxygen, rather than directly measuring cerebral blood flow. (LLM, conf=0.82)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Recent studies using direct intracranial recordings have been instrumental to define a functional organization of the SMC and the precise cortical dynamics for speech movements17,24,43,44 (Fig. 1; Bouchard et al. 2013, Mugler et al. 2018, Chartier et al. 2018 and Dichter et al. 2018).”
§ The speech neuroprosthesis > Cortical encoding of speech features
What the sources say
- Supportedcites [17] source
The Abstract reports high-resolution cortical recordings during syllable production and describes both the organization of speech-articulator representations and their transitions over tens of milliseconds. The Discussion likewise says the direct recordings revealed spatial and temporal profiles of speech-articulator representations in human ventral SMC. The supplied passages do not give page numbers. (LLM, conf=0.93)
- Supportedcites [24] source
The Introduction describes high-density intracranial ECoG recordings during continuous speech to study articulatory movement dynamics. The Results (Kinematic organization of vSMC) report spatial organization by kinematic function and place of articulation, supporting the claim about functional organization. Page numbers are not available in our copy. (LLM, conf=0.91)
- Not matchedcites [43]
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Supportedcites [44] source
The Introduction describes high-density intracranial recordings used to resolve rapid pitch changes in natural speech and identify the functional roles of laryngeal motor cortex regions. The Discussion says their spatial and temporal resolution revealed local encoding in those regions at the time scale of speech intonation, supporting this paper’s contribution to the cited group of studies. (LLM, conf=0.90)
boosteddirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Previously, data on the long-term safety and stability of ECoG primarily came from its application in epilepsy82,90–92, including the paddle-style surface strip electrodes with the Neuropace responsive neurostimulation system, which has demonstrated safety and signal stability for at least 9 years92.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Neural-recording interfaces
What the sources say
- Partly supportedcites [82] sourceExperimental — accuracy not yet measured
Split into 2 atoms — 0 confirmed, 2 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.
- Not checked — no full text of the sourcecites [90] source
- Not matchedcites [91]
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Partly supportedcites [92] sourceExperimental — accuracy not yet measured
Split into 2 atoms — 0 confirmed, 2 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.
boostedcorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Converging evidence suggests that Broca’s area, although classically posited as critical for speech production, may not have a critical role in speech articulation15,69,70 but may instead be involved in word retrieval and grammar71,72, although this is an active research area71.”
§ The speech neuroprosthesis > Cortical encoding of speech features
What the sources say
- Supportedcites [15] source
The Main section reports little to no information about orofacial movements, phonemes or words in area 44, which it identifies as part of Broca’s area. It says this finding is consistent with work questioning Broca’s area’s traditional role in speech, supporting the citing sentence’s hedged articulation claim. The word-retrieval and grammar clause has different citations and is not attributed to this paper. (LLM, conf=0.85)
- Partly supportedcites [69] sourceExperimental — accuracy not yet measured
The Conclusions (page not provided) report that resection of anatomically defined Broca’s area was not significantly associated with Broca’s aphasia or speech fluency in this cohort. That supports questioning its critical role in fluent speech, but the study did not directly test speech articulation, which is more specific. The word-retrieval and grammar clause has separate citations and is not attributed to this source. (LLM, conf=0.82)
- Not checked — no full text of the sourcecites [70] source
- Partly supportedcites [71] sourceExperimental — accuracy not yet measured
The “Posterior inferior frontal gyrus” section links Broca’s area, particularly its anterior sector, to grammatical aspects of speech production. Our copy does not establish a role for Broca’s area in word retrieval; searches for “word retrieval” and “lexical retrieval” found no supporting passage. The separate articulation claim is attached to other citations in the citing sentence. (LLM, conf=0.85)
- Partly supportedcites [72] sourceExperimental — accuracy not yet measured
The Results and Discussion (pages unavailable in our copy) support a role for the left inferior frontal gyrus, including Broca’s area, in syntactic encoding and grammar. They do not establish a role in word retrieval; the study distinguished its syntactic-encoding sites from sites that disrupted picture naming. The articulation clause has separate citations and is not assessed here. (LLM, conf=0.82)
hedgedconditional- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Successful classification of a limited set of produced sentences in the setting of question-and-answer dialogue has been demonstrated, where incorporating context from the predicted question improved classification of the answer117.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Partly supportedcites [117] sourceExperimental — accuracy not yet measured
The Methods (Experimental task design) describe a limited question-and-answer task with 24 possible spoken answers, and the Results (Question and answer decoding performance) report higher answer-classification accuracy when predicted-question context was integrated. But Table 1 shows that the produced answers were mostly single words or short phrases, not sentences; the questions were the sentences participants heard. (LLM, conf=0.87)
boosteddirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Partly supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Individuals with dysarthria may retain some intelligible vocalization over single, isolated words, allowing the acoustic regression approach to prove successful in synthesizing single words13 (Fig. 1; Angrick et al. 2023; and Supplementary Table 1).”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Supportedcites [13] source
The Introduction says the participant with ALS could produce individual words that were intelligible when spoken in isolation, although his sentences were often unintelligible. It also describes mapping his neural activity to acoustic features; the Discussion reports successful synthesis of intelligible words from a closed vocabulary. (LLM, conf=0.92)
- Supportedcites Angrick et al. 2023 source
The Introduction reports that the participant, who had substantial speech impairment, could produce isolated words intelligibly, though his sentences were often unintelligible. It also describes mapping his neural activity to acoustic features; the Discussion reports successful synthesis of intelligible words from a closed vocabulary. The claim describes this possibility rather than asserting that it applies to all individuals with dysarthria. (LLM, conf=0.90)
hedgedconditional- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“In this approach, MEA activity is decoded into subword units and converted into sentences using language models; this language model output is used as a ‘pseudo-label’ to update the RNN decoding model9,16,105.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Supportedcites [9] source
The 'Decoding handwritten sentences' section says the RNN converts neural activity into character probabilities and a language model processes the output. The 'Daily decoder retraining' section says the language model error-corrects the output and is used to retrain the decoder automatically, supporting the description of its output as pseudo-labels. The paper describes character-level decoding specifically; it does not use the term 'subword units.' (LLM, conf=0.86)
- Partly supportedcites [16] sourceExperimental — accuracy not yet measured
Methods: Decoding speech describes neural activity being decoded into phonemes, then combined into words and sentences using language models. It does not say that language-model output serves as pseudo-labels to update an RNN. The Discussion instead describes rapid recalibration using ground-truth labels from displayed Copy Task sentences; the claimed automatic update mechanism is not established by these passages. (LLM, conf=0.86)
- Supportedcites [105] source
Section 3.1 (page unavailable) says neural activity from microelectrode arrays is decoded by a GRU RNN into character probabilities, which language models process into words and a final sentence. The Introduction and §3.2.1 say the language-model-corrected output is used as a pseudo-label to retrain the decoder. The paper specifies characters as the subword units. (LLM, conf=0.90)
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Anumanchipalli et al. added a first step of decoding an articulatory representation from neural activity, which was then mapped to the intermediate acoustic representation and vocoded into speech110.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Supportedcites [110] source
The Speech decoder design section describes decoding articulatory features from neural activity, then acoustic features from those articulatory features, before synthesizing an audio signal. The Methods section, “Speech synthesis from acoustic features,” specifies that the decoded acoustic features were used to generate speech waveforms. (LLM, conf=0.93)
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Metzger et al. used this approach to decode neural activity during silent speech attempts from a participant with anarthria into synthesized speech; however, rather than regressing the mel-spectrogram, they decoded input sequences of neural activity into output sequences of discrete acoustic-speech units14.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Supportedcites [14] source
The Methods (Participant; page unavailable) identify the participant as having anarthria. The Speech synthesis and Methods (Speech synthesis; pages unavailable) describe decoding neural activity during silent speech attempts using CTC loss to predict discrete speech units, which were then converted into audible speech. The neural decoder predicted units rather than directly regressing a mel spectrogram; a downstream model generated the spectrogram. (LLM, conf=0.94)
plaindirectSingle case- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Finally, an individual may feel greater embodiment when using a speech neuroprosthesis with a personalized voice that reflects their likeness14.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Supportedcites [14] source
The Abstract says that speech’s expressiveness and identity can enhance embodied communication, and the Speech synthesis section describes a voice personalized from the participant’s pre-injury speech. Together, these passages support the citing sentence’s hedged suggestion, though the paper does not report measuring a feeling of embodiment. (LLM, conf=0.85)
hedgedconditional- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“As a first approach to counter a lack of alignment between neural activity and ground-truth targets, Moses et al.12 instructed a participant with anarthria to attempt to speak sentences word-by-word (with pauses between words).”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Partly supportedcites [12] sourceExperimental — accuracy not yet measured
The Methods (Word and Sentence Task Design; page not supplied) say the participant with anarthria attempted the words of each sentence in order. The Discussion identifies a time-alignment problem, but the Methods do not say he was instructed to pause between words; they say he attempted them at the fastest speed he could perform comfortably. (LLM, conf=0.85)
plaindirectSingle case- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“For example, brainstem-stroke survivors4,133,134 and individuals with ALS135 can retain varying degrees of control over orofacial movements and vocalization.”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Evaluation and standardization
What the sources say
- Partly supportedcites [4] sourceExperimental — accuracy not yet measured
The Results (Table 2; page unavailable) report remaining voluntary mouth, face, or head movements in 25 of 42 patients with locked-in syndrome, and the Introduction identifies stroke and ALS as possible causes. The paper does not report retained control of vocalization; the Discussion describes the patients sought for the study as severely paralyzed people with no speech. (LLM, conf=0.86)
- Partly supportedcites [133] sourceExperimental — accuracy not yet measured
The Results section, “Residual movement and current communication channels” (page unavailable), reports residual head-or-mouth movement and residual speech, including sounds or short words, in the neuromuscular-disorder group. The Methods section, “Participants” (page unavailable), says that group included ALS alongside other disorders. The paper does not report these findings separately for participants with ALS, so it does not establish the ALS-specific part of the claim. (LLM, conf=0.82)
- Not checked — no full text of the sourcecites [134] source
- Not checked — no full text of the sourcecites [135] source
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Partly supported
The cited sources that could be checked couldn’t confirm or dispute this claim.
In the paper
“however, for individuals with vocal-tract paralysis, their maximal rate of attempted speech can vary considerably and may depend on their residual articulatory capacities and degree of comfort4,133,134.”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Evaluation and standardization
What the sources say
- Could not verifycites [4] sourceExperimental — accuracy not yet measured
The Results and Discussion (pages not available in our copy) describe variation in remaining voluntary movements and communication methods among patients with locked-in syndrome. They do not address the maximal rate of attempted speech or whether it depends on residual articulatory capacity or comfort; searches for those terms found no pertinent passage. (LLM, conf=0.75)
- Could not verifycites [133] sourceExperimental — accuracy not yet measured
The Results sections report differences in residual speech and preferences for attempted speech, but do not measure maximal attempted-speech rate or link it to articulatory capacity or comfort. The questionnaire methods explicitly focused on mental-strategy preferences rather than system speed. Searches for speech rate, articulation, and comfort found no passage addressing the claim. (LLM, conf=0.76)
- Not checked — no full text of the sourcecites [134] source
hedgedconditional- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Could not verify
Every source the paper cites for this claim was checked and supports it.
In the paper
“Able speakers often speak at rates of around 150 WPM during conversational speech10; however, for individuals with vocal-tract paralysis, their maximal rate of attempted speech can vary considerably and may depend on their residual articulatory capacities and degree of comfort4,133,134.”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Evaluation and standardization
What the sources say
- Supportedcites [10] source
In “Toward a Speech Neuroprosthesis” (page not specified), the paper gives natural speaking rates of 120–150 words per minute in healthy speakers. This supports the cited estimate of around 150 WPM for able speakers. (LLM, conf=0.91)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Importantly, these large language models have the capacity to integrate conversational context into their predictions, which has been shown to improve neural-decoding pipelines117.”
§ The speech neuroprosthesis > Future directions
What the sources say
- Partly supportedcites [117] sourceExperimental — accuracy not yet measured
The Discussion reports that using the predicted question as conversational context improved decoding of the subsequent answer. The Methods > Modeling describes a context-integration model built from question likelihoods and predefined answer priors, not a large language model. The paper therefore supports the improvement from context, but not the attribution of that result to large language models. (LLM, conf=0.88)
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Finally, speech decoders have yet to fully leverage paradigms of feedback-based learning that have the potential to entrain new patterns of neural activity for better control171–174.”
§ The speech neuroprosthesis > Future directions
What the sources say
- Could not verifycites [171] sourceExperimental — accuracy not yet measured
The Introduction and Experimental Paradigm describe intracranial recordings during speech initiation and a picture-naming task, not speech-decoder training. Searches for feedback-based learning, entrainment, and decoders found no passage addressing whether such paradigms have been underused or could improve decoder control. Page numbers were not provided in our copy. (LLM, conf=0.75)
- Partly supportedcites [172] sourceExperimental — accuracy not yet measured
The Discussion (page not provided) reports that new neural activity patterns emerged during learning of a monkey cursor-control BCI and led to improved behavior. It also suggests that guiding such patterns might facilitate learning in other contexts. The paper does not assess speech decoders or say whether they have fully leveraged feedback-based learning. (LLM, conf=0.87)
- Partly supportedcites [173] sourceExperimental — accuracy not yet measured
The Discussion (page unavailable) reports that closed-loop feedback during walking-avatar control changed cortical activity and could support adaptation learning and better BCI control. It does not address whether speech decoders have underused feedback-based learning, so the speech-specific part of the claim is unsupported. (LLM, conf=0.86)
- Partly supportedcites [174] sourceExperimental — accuracy not yet measured
Section 2 (page unavailable) says that neurofeedback during motor imagery can promote changes in neural activation and help users generate signals detectable by a BCI classifier. That supports the proposed benefit of feedback-based learning, but the paper does not establish that speech decoders have yet to fully use such paradigms; its discussion concerns motor-imagery BCIs. (LLM, conf=0.86)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
- Could not verify
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“In addition, although not ready for chronic implantation, NeuroPixels, or similar probes that sample laminar cortex, may offer a method to record high-density single units across cortical layers in humans187,188.”
§ The speech neuroprosthesis > Future directions
What the sources say
- Partly supportedcites [187] sourceExperimental — accuracy not yet measured
The Electrophysiology figure captions describe recordings across cortical layers in rats, and the Discussion describes a high-channel-count, single-spike-resolution system as a prototype for future human implants. The Introduction says no clinically translatable microelectrode technology for large-scale recordings was then available. The paper does not establish that NeuroPixels specifically could provide laminar single-unit recordings in humans; its reported recordings used Neuralink polymer probes in rats. (LLM, conf=0.86)
- Supportedcites [188] source
The Abstract reports simultaneous recordings from over 200 well-isolated cortical single units in humans, and the Discussion says the probe resolves units at different cortical depths. The study describes acute recordings and calls its approach a step toward developing chronic neural interfaces, consistent with the qualification that these probes were not yet ready for chronic implantation. (LLM, conf=0.88)
hedgedconditional- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“A recent non-peer-reviewed preprint found that neural populations in the SMA are among the earliest predictors of speech onset and may remain strongly activated in imagined and attempted speech170.”
§ The speech neuroprosthesis > Future directions
What the sources say
- Partly supportedcites [170] sourceExperimental — accuracy not yet measured
The fMRI Paradigms section describes covert word-generation tasks, and the Figure 2 caption reports pre-SMA activation during those tasks. But this Journal of Neurosurgery case report does not establish that SMA neural populations are among the earliest predictors of speech onset or that they remain strongly activated during attempted speech; it is also not a non-peer-reviewed preprint. (LLM, conf=0.84)
hedgedcorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Similarly, both in academic and industry settings, approaches are being developed to improve throughput, sampling, biocompatibility and stability of MEAs183–186.”
§ The speech neuroprosthesis > Future directions
What the sources say
- Partly supportedcites [183] sourceExperimental — accuracy not yet measured
The Methods (Device components; page not provided) describe an industry-developed, implantable 16-sensor electrode array, and the Results (Preliminary safety reporting; page not provided) report no device-related adverse events in two participants. This makes the paper an example of a developing array approach, but our copy does not show that it improves MEA throughput or sampling, or establish improved biocompatibility or long-term stability. (LLM, conf=0.82)
- Not checked — no full text of the sourcecites [184] source
- Supportedcites [185] source
The Spatial integration section (page unavailable) discusses increasing recorded-neuron counts and sampling resolution in MEAs. The Encouraging novel technologies and Stable long-term recording sections (pages unavailable) describe efforts to reduce tissue response and improve recording stability. Together, these passages support the attributed claim that approaches are being developed to improve MEA throughput, sampling, biocompatibility and stability. (LLM, conf=0.90)
- Supportedcites [186] source
The Discussion (no page provided) describes a microwire-CMOS array scalable to tens of thousands of channels and a next-generation device already in development for greater electrode density and reduced tissue damage. It also describes the robustness and longevity of microwires; the Introduction (no page provided) discusses their suitability for chronic use and the potential to reduce foreign-body response. Together, these support the broad claim that approaches are being developed to improve MEA sampling, throughput, biocompatibility, and stability. (LLM, conf=0.87)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Transfer learning, either between individuals108 or different types of speech within the same individual120, may also be used to expedite decoder training.”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Practical implementation of a speech neuroprosthesis
What the sources say
- Partly supportedcites [108] sourceExperimental — accuracy not yet measured
The Results section (‘Transfer learning’; page not provided) reports that adding other sentences spoken by the same participant improved decoding of the target sentence set. This supports the cited point that related speech data may help decoder training, but the paper tests different sentence sets, not clearly different types of speech. The between-individual claim is attached to a different citation. (LLM, conf=0.82)
- Partly supportedcites [120] sourceExperimental — accuracy not yet measured
The section “Distinctions in evoked neural activity between silent- and overt-speech attempts” reports pretraining on one speech type and fine-tuning on the other within the same participant, with improved classification accuracy. It does not report whether this reduced decoder training time or the amount of training data needed. The separate between-individual claim is attached to another citation, not this paper. (LLM, conf=0.82)
hedgedconditional- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Different approaches to increase the density and lower the invasiveness of ECoG are currently being developed for use in humans and animal models176–182.”
§ The speech neuroprosthesis > Future directions
What the sources say
- Partly supportedcites [176] sourceExperimental — accuracy not yet measured
The Introduction and Discussion (page numbers unavailable) describe a high-density thin-film ECoG array tested in humans and note related arrays used in animal studies. The paper describes the arrays as safe and conformal, but does not establish that the human approach lowers ECoG invasiveness; that part of the claim is not supported by this source. (LLM, conf=0.83)
- Supportedcites [177] source
The Results describe a high-density µECoG array used in human patients, and the Methods (Intra-operative µECoG array implantation; page not provided) explain that its narrow, flexible design can be inserted through a burr hole opened for another procedure. The Introduction also discusses high-density µECoG work in rodents and non-human primates. Together, these passages support the cited paper as an example of the approaches described. (LLM, conf=0.86)
- Partly supportedcites [178] sourceExperimental — accuracy not yet measured
The Introduction describes efforts to increase ECoG electrode density, and the Discussion reports this paper’s 1152-electrode array tested in a monkey. The Introduction also describes ECoG as relatively low-invasive and mentions human and nonhuman-primate applications, but it does not say that this approach lowers ECoG’s invasiveness further. (LLM, conf=0.84)
- Supportedcites [179] source
The Introduction describes μECoG arrays as high-resolution interfaces that are less invasive to cortical tissue than penetrating microelectrode arrays. It also reports their use in rats, non-human primates, and humans undergoing epilepsy monitoring, supporting this paper as an example of the approaches cited. (LLM, conf=0.91)
- Supportedcites [180] source
The Conclusion (page unavailable) describes a high-density μECoG array that uses multiplexing to reduce output wires and the device’s footprint at the implantation site. It reports recordings over the rat auditory cortex, making this paper an example of the approaches cited for animal models. The sentence’s multiple citations do not require this paper alone to establish human use. (LLM, conf=0.89)
- Supportedcites [181] source
The Introduction describes minimally invasive, high-density µECoG arrays used in human patients, and the Discussion contrasts their flexible surface design with penetrating arrays. This paper supplies a human example of the approaches described; the sentence’s other citations need not each cover animal models. (LLM, conf=0.90)
- Supportedcites [182] source
The Introduction describes a high-density, minimally invasive µECoG platform intended to facilitate human clinical use. The Minimally Invasive Surgical Implantation section reports testing in Göttingen minipigs, supporting this paper as an example of an approach developed for human use and studied in an animal model; it does not claim human testing. (LLM, conf=0.91)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“The supplementary motor area (SMA) is another important region in the speech production network165,169 that is involved in speech initiation170.”
§ The speech neuroprosthesis > Future directions
What the sources say
- Could not verifycites [165] sourceExperimental — accuracy not yet measured
The Introduction and Methods describe hand-movement representations in primary motor and sensory cortex, not the supplementary motor area’s role in speech. Searches for the supplementary motor area, speech production, and speech initiation found no passage addressing the attributed claim. (LLM, conf=0.76)
- Could not verifycites [169] sourceExperimental — accuracy not yet measured
The Discussion identifies STG, SMG, and dorsal precentral gyrus as speech-network regions, but does not identify the supplementary motor area (SMA) or assign it a role in speech initiation. Searches for “supplementary motor area,” “SMA,” and “speech initiation” found no passage addressing the attributed claim. No page numbers were provided for the indexed passages. (LLM, conf=0.76)
- Supportedcites [170] source
The introduction describes speech hesitancy through mutism after surgery near the language-dominant SMA, and the Discussion reports pre-SMA activation during language tasks. Together, these passages support the SMA region’s role in speech production and initiation; the Discussion also links a tract connecting the pre-SMA with Broca’s area to speech-initiation deficits. Page numbers were not available in our copy. (LLM, conf=0.86)
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Could not verify
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Separately, the feasibility of non-invasive methods for communication brain computer interfaces is being actively investigated77–80, with the goal of improving signal-to-noise and spatial resolution.”
§ The speech neuroprosthesis > Future directions
What the sources say
- Partly supportedcites [77] sourceExperimental — accuracy not yet measured
The Discussion (page not supplied) reports non-invasive functional ultrasound recording through a previously implanted acoustic window and describes task-state decoding as an initial step toward human brain-machine interfaces. The Introduction and Conclusion (pages not supplied) discuss the need for better sensitivity and spatial precision. The study does not investigate a communication interface, and its recording method requires prior skull reconstruction, so it does not establish a wholly non-invasive method. (LLM, conf=0.85)
- Partly supportedcites [78] sourceExperimental — accuracy not yet measured
The Introduction describes non-invasive EEG and fNIRS as methods used for communication BCIs and says hybrid systems are studied to improve classification accuracy and information transfer rate. It does not identify improved signal-to-noise ratio or spatial resolution as the goal; the paper instead focuses on maintaining performance with fewer sensors. The indexed passages provide section labels but no page numbers. (LLM, conf=0.87)
- Not checked — no full text of the sourcecites [79] source
- Partly supportedcites [80] sourceExperimental — accuracy not yet measured
The Introduction (page unavailable) reports a noninvasive BCI used by locked-in patients to select letters for communication. The “Ultrasound Brain Functional Imaging–reading” section (page unavailable) discusses low spatial resolution in EEG, reduced signal-to-noise ratio for deep MEG sources, and advances in ultrasound resolution. It does not specifically say that improving both measures is the goal of research on communication BCIs. (LLM, conf=0.83)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
- Partly supported
In the paper
“Pitch can be modulated over the course of a phrase to convey additional meaning to the produced words (phrasal prosody), such as whether an utterance is a statement or question (rise in pitch at the end of the phrase)11.”
§ The speech neuroprosthesis > Overview of speech features
What the sources say
- Not matchedcites [11]
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
- Not matched
In the paper
“In addition, pitch is an important aspect of the uniqueness and expressivity of the voice of a person31.”
§ The speech neuroprosthesis > Overview of speech features
What the sources say
- Not checked — no full text of the sourcecites [31] source
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
In the paper
“Often, a mel-frequency scale — in which frequency is logarithmically transformed — is used to visualize the speech spectrogram, as this emphasizes power in perceptually salient frequency bands that better correspond to how the human auditory system processes sound27 (Fig. 2e).”
§ The speech neuroprosthesis > Overview of speech features
What the sources say
- Not checked — no full text of the sourcecites [27] source
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
In the paper
“Speech neuroprostheses have the potential to not only enable more natural communication of words and sentences but also restore other expressive components of communication that convey meaning, such as intonation, loudness and facial gestures11.”
§ The speech neuroprosthesis > Introduction
What the sources say
- Not matchedcites [11]
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
hedgedconditional- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Not matched
In the paper
“During continuous speech, the vocal-tract muscles move rapidly between different configurations that may be grouped as labial, front tongue, back tongue and vocalic21 (Fig. 2b) on the basis of where air is constricted in the vocal tract.”
§ The speech neuroprosthesis > Overview of speech features
What the sources say
- Not matchedcites [21]
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
hedged- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
- Not matched
In the paper
“In tonal languages, such as Mandarin, the meaning of individual words depends on their associated pitch contour (lexical tone)30.”
§ The speech neuroprosthesis > Overview of speech features
What the sources say
- Not checked — no full text of the sourcecites [30] source
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
In the paper
“The envelope of the acoustic waveform illustrates the intensity of speech over time (Fig. 2d), which is an important measurement that correlates with speech rate, global stress patterns and loudness11.”
§ The speech neuroprosthesis > Overview of speech features
What the sources say
- Not matchedcites [11]
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
- Not matched
In the paper
“These articulator movements are coordinated with expiration of air to produce the sound waves of speech19.”
§ The speech neuroprosthesis > Overview of speech features
What the sources say
- Not checked — no full text of the sourcecites [19] source
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
None of the sources cited for this claim could be matched to a paper in Crossref, so there is nothing to check it against.
In the paper
“Phonemes are defined as the smallest perceptually distinct units of sound that form a language (39 unstressed phonemes exist in English)32,33.”
§ The speech neuroprosthesis > Overview of speech features
What the sources say
- Not matchedcites [32]
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Not matchedcites [33]
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
- Not matched
In the paper
“Phonemes are particularly useful for text decoding, given that they form a discrete, low-dimensional representation of speech (~39D in English) that can be scaled to produce the much larger vocabularies of words and sentences21.”
§ The speech neuroprosthesis > Overview of speech features
What the sources say
- Not matchedcites [21]
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Not matched
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“circumventing motor-pathway injury by directly decoding speech from intact cortical activity has the potential to restore natural communication and self-expression”
§ The speech neuroprosthesis > Abstract
hedgedconditional- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
In the paper
“Phonemes can be grouped on the basis of similarities in their sound patterns, and their articulatory features — the place of articulation of a phoneme in the vocal tract — provide one important grouping category21.”
§ The speech neuroprosthesis > Overview of speech features
What the sources say
- Not matchedcites [21]
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
plaincorrelational- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
- Not matched
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Thus, although the ventral–dorsal axis of the SMC issomatotopically arranged in brain areas at the level of neural populations, at the single-neuron level, it may have a more heterogeneous and distributed arrangement.”
§ The speech neuroprosthesis > Cortical encoding of speech features
hedged- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Although these non-invasive techniques have facilitated decoding of limited vocabularies of imagined words, syllables and speech sounds75,76 and reconstruction of semantic content of speech68 (Fig. 1; Tang et al. 2023), they are not portable — with the exception of EEG — and low spatial or temporal resolution has hindered scaling decoding to larger vocabularies and longer speech segments (senten”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Neural-recording interfaces
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Although more work is needed to validate long-term robustness and signal stability with MEAs, their potential for high-performance speech decoding in combination with computational recalibration for stable acquisition offers potential for future use as a clinically viable neuroprosthesis.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Neural-recording interfaces
hedgedconditional- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
In the paper
“These techniques have relatively high temporal resolution but lack spatial specificity, with each sensor recording from a large area of the brain73.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Neural-recording interfaces
What the sources say
- Not checked — no full text of the sourcecites [73] source
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
None of the sources cited for this claim have been checked yet.
In the paper
“By targeting words within sentences, rather than whole sentences, and leveraging contemporary machine-learning techniques, this RNN-based approach improved accuracy and the overall vocabulary size and number of sentences that could be decoded.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
None of the sources cited for this claim have been checked yet.
In the paper
“A speech-detection model used cortical activity, primarily from the SMC, to predict when the individual was attempting to speak a word and passed the corresponding neural features to a word-classification model trained to predict probabilities across 50 words in a predefined vocabulary.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Although this approach facilitated alignment-free speech synthesis, suited to scale to fully locked in individuals, it is not yet ideal for low-latency streaming.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“An advantage of this concatenative-synthesis approach is its feasibility with smaller dataset sizes; however, it does not fully leverage advances in machine learning, relying on time window correlations.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“As spoken language is a more fundamental form of human communication than written language (text), this approach can enable more fine-grained and natural control over decoded outputs.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
plainconditional- Asserted: passes — The authors put it forward as true.
- Checkable: fails — It may be too vague to check against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. It may be too vague to check against evidence.
- Rests on an assumption: Spoken language is a more fundamental form of human communication than written language.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“This acoustic regression approach achieved higher-performance speech synthesis, probably owing to the ability of a deep-neural network to learn complex nonlinear mappings between inputs and outputs.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
hedgedconditional- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“However, these potential benefits come with increased difficulty; speech synthesis has generally proven more challenging than text decoding, as it does not leverage language models and predefined vocabularies.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
The paper states this without a source or data behind it here, so there is nothing to check it against. The quote shows it is in the paper.
In the paper
“Although this approach facilitated access to a large vocabulary, while still using intended speech, decoding speeds were slower and communication was less natural for the individual.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
In the paper
“It is possible that low-latency restoration of audible speech could have analogous benefits to rapid closed-loop feedback for neuroprosthetic control in other motor domains124.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Not checked — no full text of the sourcecites [124] source
hedgedconditional- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Although the speech-detection model used provided word-level alignment, direct classification from a 50-word vocabulary limited this approach.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
In the paper
“The individual then had the ability to spell character-by-character and language models could transform decoded-character sequences into sentences comprised from a vocabulary of more than 1,000 words (Supplementary Table 1).”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Not matchedcites Supplementary Table 1
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
plaindirectSingle case- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
- Not matched
A claim is checked only against the cited paper’s full text, never its abstract, and none is available for its sources.
In the paper
“Often in ASR, to prevent equal weighing of short and long sentences, error rates are computed over paragraphs or longer segments of text143,144 rather than across individual sentences.”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Evaluation and standardization
What the sources say
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
In the paper
“For example, in English, the most common 1,000 words can cover more than 85% of the content in spoken sentences156.”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Evaluation and standardization
What the sources say
- Not checked — no full text of the sourcecites [156] source
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
A claim is checked only against the cited paper’s full text, never its abstract, and none is available for its sources.
In the paper
“In ASR, a 5% WER, over longer segments of text, is standard for professional-level transcriptions141 and 25% or less is considered acceptable142.”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Evaluation and standardization
What the sources say
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Open-ended human-transcribed WER complements MCD by providing a more interpretable metric derived from the perceptual intelligibility of the synthesized speech.”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Evaluation and standardization
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim.
In the paper
“Most previous neural speech-synthesis systems have not been highly intelligible with an open-ended vocabulary152 and, thus, have used forced-choice listening tests or word-level classification accuracies instead of open-ended WER.”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Evaluation and standardization
What the sources say
- Not checked — no full text of the sourcecites [152] source
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
It may state two separate claims as one.
None of the sources cited for this claim have been checked yet.
In the paper
“Non-verbal facial expressions that accompany virtual or face-to-face communication provide numerous benefits compared with audio-only communication, including improved conveyance of emotion and attitude130,131, that provides increased clarity132.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
What the sources say
- Not matchedcites [130]
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Not matchedcites [131]
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Not checked — no full text of the sourcecites [132] source
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
- Not matched
In the paper
“However, lower frequency words remain an important component of natural language because they facilitate expressivity156.”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Evaluation and standardization
What the sources say
- Not checked — no full text of the sourcecites [156] source
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: fails — It may be too vague to check against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. It may be too vague to check against evidence. The paper’s conclusions may not depend on it.
In the paper
“A language’s vocabulary is characterized by Zipf’s law, from which it follows that the most common words in a language make up a disproportionate amount of spoken content in that language155.”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Evaluation and standardization
What the sources say
- Not checked — no full text of the sourcecites [155] source
plaindirect- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“For large-vocabulary sizes, it is not feasible to evaluate performance on every word.”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Evaluation and standardization
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
In the paper
“A growing literature aims to define general machine-learning principles to encourage the development of scalable, reliable and efficient systems160.”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Practical implementation of a speech neuroprosthesis
What the sources say
- Not checked — no full text of the sourcecites [160] source
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Specifically, high-performance speech decoding has yet to be demonstrated in an individual that is fully locked-in with negligible residual motor function.”
§ The speech neuroprosthesis > Future directions
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: fails — The paper may give no reason why it holds.
Probably not a substantial claim. The paper may give no reason why it holds.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Importantly, many of the aforementioned user concerns — ease of use, re-training periods and data privacy — are not unique to neuroprostheses but rather apply to many machine-learning-based applications.”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Practical implementation of a speech neuroprosthesis
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
In the paper
“For example, in some situations, users may prefer to limit their output space of speech to single, high-utility words and/or phrases (such as ‘no’, ‘yes’, ‘thank you’, ‘not now’ or ‘bring that’) in exchange for near-perfect decoding performance157.”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Practical implementation of a speech neuroprosthesis
What the sources say
- Not checked — no full text of the sourcecites [157] source
hedged- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it.
A conclusion the paper draws from its own results — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Although spelling and classification approaches may be critical to a practical system for daily use, they do not address users’ desire for high-speed and natural communication.”
§ The speech neuroprosthesis > Best practices for evaluating and implementing speech neuroprostheses > Practical implementation of a speech neuroprosthesis
hedged- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: fails — The paper’s conclusions may not depend on it.
- Based: passes — The paper gives a reason why it holds.
Probably not a substantial claim. The paper’s conclusions may not depend on it. It may state two separate claims as one.
The paper relies on this without support.
In individuals with dysarthria or anarthria, cerebral cortex neuronal populations that contain representations of articulatory movements may remain intact.
“Importantly, in individuals with dysarthria or anarthria, cerebral cortex neuronal populations, which contain representations of articulatory movements used to produce sounds, may remain intact.”
The paper relies on this without support.
Spoken language is a more fundamental form of human communication than written language.
“As spoken language is a more fundamental form of human communication than written language (text), this approach can enable more fine-grained and natural control over decoded outputs.”
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
“Text (Fig. 3b) and synthesized speech (Fig. 3c) are two outputs commonly used in speech-decoding systems.”
§ The speech neuroprosthesis > Decoding speech from articulatory features > Modelling speech features from neural activity
“In the second stage, a speech synthesizer — a vocoder — is used to convert the acoustic representation into an audible speech waveform (Fig. 3b).”
Citations
Cites
“Librispeech: an ASR corpus based on public domain audio books,”
Panayotov, V., Chen, G., Povey, D., & Khudanpur, S. (2015) · 2015 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) · doi:10.1109/icassp.2015.7178964 · Cited by 6,072
Graves, A., Fernández, S., Gomez, F., & Schmidhuber, J. (2006) · Proceedings of the 23rd international conference on Machine learning - ICML '06 · doi:10.1145/1143844.1143891 · Cited by 5,370
LLaMA: Open and Efficient Foundation Language Models
Touvron, H., Lavril, T., Izacard, G., Martinet, X., Lachaux, M.-A., Lacroix, T., Rozière, B., Goyal, N., Hambro, E., Azhar, F., et al. (2023) · arXiv · doi:10.48550/arxiv.2302.13971 · Cited by 4,602 · cited 1×
“Another approach to improve decoding performance is to leverage large language models, which are currently being developed in industry 145,195 , to score the likelihood of decoded sentences from neural activity.”
PENFIELD, W., & BOLDREY, E. (1937) · Brain · doi:10.1093/brain/60.4.389 · Cited by 4,188
An Introduction to Variational Autoencoders
Kingma, D. P., & Welling, M. (2019) · Foundations and Trends® in Machine Learning · doi:10.1561/2200000056 · Cited by 2,609 · cited 1×
“One potential avenue is to leverage advancements in self-supervised learning 189 to learn latent representations that best capture behaviour-related dynamics, even when noise is present in the neural activity 121,190–194 .”
Visual Contribution to Speech Intelligibility in Noise
Sumby, W. H., & Pollack, I. (1954) · The Journal of the Acoustical Society of America · doi:10.1121/1.1907309 · Cited by 2,583 · cited 1×
“Non-verbal facial expressions that accompany virtual or face-to-face communication provide numerous benefits compared with audio-only communication, including improved conveyance of emotion and attitude 130,131 , that provides increased clarity 132 .”
SWITCHBOARD: telephone speech corpus for research and development
Godfrey, J. J., Holliman, E. C., & McDaniel, J. (1992) · [Proceedings] ICASSP-92: 1992 IEEE International Conference on Acoustics, Speech, and Signal Processing · doi:10.1109/icassp.1992.225858 · Cited by 2,134
Robust Speech Recognition via Large-Scale Weak Supervision
Radford, A., Kim, J. W., Xu, T., Brockman, G., McLeavey, C., & Sutskever, I. (2022) · arXiv · doi:10.48550/arxiv.2212.04356 · Cited by 1,953
Sequence Transduction with Recurrent Neural NetworksPreprint
Graves, A. (2012) · arXiv (Cornell University) · doi:10.48550/arxiv.1211.3711 · Cited by 1,842 · cited 1×
“Contemporary machine-learning architectures, tailored for streaming, may also reduce the latency at which speech-synthesis and text-decoding systems operate 196,197 .”
Natural speech reveals the semantic maps that tile human cerebral cortex
Huth, A. G., de Heer, W. A., Griffiths, T. L., Theunissen, F. E., & Gallant, J. L. (2016) · Nature · doi:10.1038/nature17637 · Cited by 1,504
Research works printed but not linked: 15
- Das, J. M., Anosike, K. & Asuncion, R. M. D. Locked-in syndrome. StatPearls https://www.ncbi.nlm.nih.gov/books/NBK559026/ (StatPearls, 2021).
- Bull, P. & Frederikson, L. in Companion Encyclopedia of Psychology (Routledge, 1994).
- Baddeley, A. Working Memory xi, 289 (Clarendon/Oxford Univ. Press, 1986).
- Penfield, W. & Roberts, L. Speech and Brain-Mechanisms (Princeton Univ. Press, 1959). This study provided insights into cortical control of speech and language through neurosurgical cases, including cortical resection, direct-cortical stimulation and seizure mapping.
- Acharya, A. B. & Maani, C. V. Conduction aphasia. StatPearls https://www.ncbi.nlm.nih.gov/books/NBK537006/ (StatPearls, 2023).
- Fernández-de Thomas, R. J., Munakomi, S. & De Jesus, O. Craniotomy. StatPearls https://www.ncbi.nlm.nih.gov/books/NBK560922/ (StatPearls, 2024).
- Jia, J., Wang, X., Wu, Z., Cai, L. & Meng, H. Modeling the correlation between modality semantics and facial expressions. In Proc. 2012 Asia Pacific Signal and Information Processing Association Annual Summit and Conference 1–10 (2012).
- Nip, I. & Roth, C. R. in Encyclopedia of Clinical Neuropsychology (eds Kreutzer, J., DeLuca, J. & Caplan, B.) (Springer International, 2017).
- Companion Volume, Short Papers 173-176 (Association for Computational Linguistics, 2007).
- Trnka, K., Mccaw, J., Mccoy, K. & Pennington, C. in Human Language Technologies 2007 173–176 (2008). · DOI · Paper page
- Cho, C. J., Chang, E. & Anumanchipalli, G. Neural latent aligner: cross-trial alignment for learning representations of complex, naturalistic neural data. In Proc. 40th Int. Conf. Machine Learning 5661–5676 (PMLR, 2023).
- Oppenheim, A. V., Schafer, R. W. & Schafer, R. W. Discrete-Time Signal Processing (Pearson, 2014).
- Duffy, J. R. Motor Speech Disorders: Substrates, Differential Diagnosis, and Management (Elsevier Health Sciences, 2019).
- Sokolov, A. Inner Speech and Thought (Springer Science & Business Media, 2012).
- Institute of Medicine (US) Roundtable on Value & Science-Driven Health Care. Clinical Data as the Basic Staple of Health Learning: Creating and Protecting a Public Good: Workshop Summary (National Academies Press, 2011).
Cited by
Sorted by citations
An instantaneous voice-synthesis neuroprosthesis
Wairagkar, M., Card, N. S., Singer-Clark, T., Hou, X., Iacobacci, C., Miller, L. M., Hochberg, L. R., Brandman, D. M., & Stavisky, S. D. (2025) · Nature · doi:10.1038/s41586-025-09127-3 · Cited by 120
A streaming brain-to-voice neuroprosthesis to restore naturalistic communication
Littlejohn, K. T., Cho, C. J., Liu, J. R., Silva, A. B., Yu, B., Anderson, V. R., Kurtz-Miott, C. M., Brosler, S., Kashyap, A., Hallinan, I. P., et al. (2025) · Nature Neuroscience · doi:10.1038/s41593-025-01905-6 · Cited by 104
“Translating these findings into speech neuroprostheses that can restore fluent communication in people with severe vocal tract paralysis has been elusive 6 .”
InfluentialSelf-citationInner speech in motor cortex and implications for speech neuroprostheses
Kunz, E. M., Abramovich Krasa, B., Kamdar, F., Avansino, D. T., Hahn, N., Yoon, S., Singh, A., Nason-Tomaszewski, S. R., Card, N. S., Jude, J. J., et al. (2025) · Cell · doi:10.1016/j.cell.2025.06.015 · Cited by 55
“Using the decoding pipeline from our previous work 12,17 (Figure 3A), we trained real-time decoders on data recorded as participants internally spoke cued sentences with their preferred inner-speech strategy (see STAR Methods).”
Restoring Speech Using Brain–Computer Interfaces
Stavisky, S. D. (2025) · Annual Review of Biomedical Engineering · doi:10.1146/annurev-bioeng-110122-012818 · Cited by 42
Emerging Medical Technologies and Their Use in Bionic Repair and Human Augmentation
Manero, A. C., Rivera, V., Fu, Q., Schwartzman, J. D., Gibbs, H. P., Shah, N., Gandhi, D. A., White, E., Crawford, K. E., & Coathup, M. J. (2024) · Bioengineering · doi:10.3390/bioengineering11070695 · Cited by 24
“Recent discoveries have developed high-spatiotemporal-resolution neural interfaces together with the adaptation of state-of-the-art speech computational algorithms that have resulted in rapid and impactful advancements in decoding neural activity into text, audible speech, and facial movements [43].”
Background
Metadata
- Authors
- Silva, A. B., Littlejohn, K. T., Liu, J. R., Moses, D. A., & Chang, E. F.
- Affiliations
- University of California, San Francisco (US)
- Corresponding author
- Chang, Edward F.
- Venue
- Nature Reviews Neuroscience
- Published
- 2024-05-14
- Volume
- 25
- Issue
- 7
- Pages
- 473-492
- ISSN
- 1471-003X (ISSN-L)
- 1471-0048
- PMC
- PMC11540306
- OpenAlex
- W4396905256
- Cited by
- 103
- References
- 207
- Citations by year
- 2026: 41 · 2025: 53 · 2024: 9
- FWCI
- 6.64
- Type
- journal article
- Language
- en
- Open access
- green
- www.ncbi.nlm.nih.gov
- Licence
- Locations
- Nature reviews. Neuroscience (publishedVersion, public-domain)
- PubMed (publishedVersion)
- PubMed Central (acceptedVersion, public-domain)
- eScholarship (California Digital Library) (submittedVersion, public-domain)
- Funders
- National Science Foundation
- National Institutes of Health
- National Institute on Deafness and Other Communication Disorders (F30DC021872, U01DC018671)
- Topics
- Life Sciences › Neuroscience › Cognitive Neuroscience › EEG and Brain-Computer Interfaces
- Life Sciences › Neuroscience › Cellular and Molecular Neuroscience › Neuroscience and Neural Engineering
- Health Sciences › Health Professions › Speech and Hearing › Voice and Speech Disorders
- Keywords
- speech decoding, motor planning, paralysis, intracranial electrodes, epilepsy monitoring, neural interfaces, augmentative and alternative communication, evaluation metrics, speech production
- MeSH terms
- Animals, Humans, Speech, Neural Prostheses, Brain-Computer Interfaces
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
A language model drew the claims from the paper on 2026-09-28, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-09-28 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
