These data are released with the manuscript: Gusman et al., 2026, Observation-Related Activity in the Human Motor Cortex Increases with Effector Anthropomorphicity.

Dataset DOI: 10.5061/dryad.37pvmcw08

Dataset Description

This dataset comprises intracortical single-unit neural activity recorded from two participants enrolled in the BrainGate2 pilot clinical trial (NCT00912041):

• T11: An ambidextrous man, in his 30s, who has tetraplegia due to a C4 AIS-B spinal cord injury that occurred over a decade prior to enrollment in the trial.
• T17: A right-handed man, in his 30s, who is completely paralyzed and anarthric due to advanced ALS.

This research was conducted under an Investigational Device Exemption granted by the US Food and Drug Administration (IDE #G090003; CAUTION: Investigational device. Limited by Federal law to investigational use) and with approval from the Institutional Review Boards of Massachusetts General Hospital (#2009P000505), the Providence VA Medical Center, and Brown University. Both participants gave informed consent.

All sessions took place at the participants' residences.

BCI behavioral tasks

• Jittered Dots Task: Participants passively observed eight visual stimuli spanning a spectrum of anthropomorphicity: a virtual Human Hand, six dot stimuli (JD0-JD5), and an Empty condition consisting of a static white fixation cross (see manuscript Fig. 1 and supplemental video S1). JD0 was created by removing the surface of the Human Hand model and placing 26 white dots at its joints, preserving the original hand kinematics. Across JD1-JD5, dot starting positions and then movement trajectories were made progressively less hand-like. Except for JD5 and Empty, each stimulus had distinct power- and pinch-grip animations. Participants fixated the screen, remained relaxed, and rated the perceived hand-likeness of each stimulus after every trial using the Anthropomorphicity Rating Scale (ARS): 4, "I definitely saw a hand"; 3, "I think I saw a hand"; 2, "I think I did NOT see a hand"; and 1, "I definitely did NOT see a hand." T11 gave verbal ratings; T17 selected ratings using eye movements. Each block contained six presentations of every effector (three per grip), and each session contained 13-16 analyzed blocks. T11 completed three sessions (post-implant trial days 385, 387, and 412), and T17 completed one session (trial day 62).
• Naive Jittered Dots Task: Before his first standard Jittered Dots session, T11 passively viewed the dot stimuli without being told that they represented hand movements (post-implant trial day 372). The first four stimulus blocks (acquisition blocks 2-5 in the distributed file) presented increasingly hand-like stimuli: Jittered, Semi-Jittered, Non-Jittered, and Human Hand. Each block contained approximately 45 pseudorandomized trials: 15 power, 15 pinch, and 15 Empty trials. During acquisition block 4 (the third stimulus block), after trial 18, T11 spontaneously reported that he had "figured out" that the dots depicted a moving hand. The distributed file also contains later post-context Active and Passive blocks, for a total of 20 retained blocks. Due to an unintended difference in the experimental software, the stimulus set used in the later blocks differed slightly from the initial set. The manuscript’s before/after analysis was therefore restricted to trials within the initial Non-Jittered block separated by T11’s spontaneous remark (the “Aha Moment”). See the manuscript and SI Appendix for additional details.
• Virtual Effector Task: Participants observed five virtual effectors: a Human Hand, an anthropomorphic Robot Hand, a three-fingered Robot Claw, a blue Cube, and an Empty condition (see manuscript Fig. 7 and supplemental video S3). The four moving effectors performed power- and pinch-grip analogues; Empty served as a no-movement baseline. Trials used an instructed-delay paradigm with an auditory grip instruction, a 587 Hz Go cue, 1.25 s of effector movement, a 0.5 s hold, a 523 Hz Stop cue, a further 0.5 s delay, and a return to neutral at three times the outward speed. Trials were presented in pseudorandomized 44-trial blocks (four repetitions of each of 11 conditions: power and pinch for each effector plus Empty/relax). In Active blocks, participants attempted and held the instructed grip between the Go and Stop cues. In Passive blocks, they only observed and were explicitly instructed not to attempt the movement. No motor-decoding feedback was provided; all trials were open loop. T11 completed three sessions (post-implant trial days 262, 276, and 280), and T17 completed one session (trial day 90).

See the manuscript and SI Appendix for complete stimulus descriptions, trial timing, instructions, trial counts, and analysis details.

Intracortical neural recordings and spike sorting

T11 had two 96-channel NeuroPort microelectrode arrays (4 mm x 4 mm, 1.5 mm electrode length, platinum tips; Blackrock Neurotech, Salt Lake City, UT) implanted in the hand-knob region of the left precentral gyrus. Both arrays were localized to cortical area 6d. Signals were analog filtered from 1 Hz to 7.8 kHz, sampled at 20 kHz, transmitted wirelessly using two Brown Wireless devices, and resampled to 30 kHz for downstream processing.

T17 had six 64-channel NeuroPort arrays (3.2 mm x 3.2 mm, 1.5 mm electrode length, iridium oxide) implanted in motor- and speech-related regions of the left precentral gyrus. Two arrays each were localized to areas 6d, 55b, and 6v. For this study, signals were recorded from the four arrays in areas 6d and 6v that modulated most strongly with attempted hand movements. Signals were analog filtered from 0.3 Hz to 7.5 kHz, sampled at 30 kHz using two NeuroPlex E devices (Blackrock Neurotech), and transmitted by cable.

Custom MATLAB and Simulink software (MathWorks, Natick, MA) stored and aligned up to 256 channels of 30 kHz neural data with task events. Single-unit activity was isolated offline using custom software that differentiated spike waveform shapes and amplitudes, then verified manually in Offline Sorter (Plexon Inc., Dallas, TX). Units recorded across different sessions may represent partially overlapping samples of the underlying neuronal population and should not be assumed to be independent neurons across days. Data from T11's lateral array in Virtual Effector Session 2 were corrupted during saving and excluded; that session contains units from the medial array only.

The distributed feat arrays contain single-unit firing rates in 20 ms bins, and the nested spikeTrains structure contains the corresponding sorted spike times and acquisition-system timing information.

Files and Variables

This dataset contains seven MATLAB v7 (.mat) files:

• sAs_JDTask_T11.mat: a 1 x 3 cell array containing T11 area 6d data from three Jittered Dots Task sessions
• sAs_JDTask_T17_6d.mat: a 1 x 1 cell array containing T17 area 6d data from one Jittered Dots Task session
• sAs_JDTask_T17_6v.mat: a 1 x 1 cell array containing T17 area 6v data from one Jittered Dots Task session
• sAs_NaiveJDTask_T11.mat: a 1 x 1 cell array containing T11 area 6d data from the Naive Jittered Dots Task session
• sAs_VETask_T11.mat: a 1 x 3 cell array containing T11 area 6d data from three Virtual Effector Task sessions
• sAs_VETask_T17_6d.mat: a 1 x 1 cell array containing T17 area 6d data from one Virtual Effector Task session
• sAs_VETask_T17_6v.mat: a 1 x 1 cell array containing T17 area 6v data from one Virtual Effector Task session

Each file contains a variable named sAs. Each cell of sAs contains one session structure (sA) with the following fields. Dimensions below use MATLAB orientation.

spikeTrains : [struct] Sorted spike times and acquisition-system timing information; see below
blks : [1 x nBlock] Block numbers included in the session structure
participantID : [char] Participant identifier ('t11' or 't17')
arrayLocationLabel : [char] Cortical-area label ('6d' or '6v')
feat : [nStep x nUnit] Firing rate of each sorted unit in 20 ms time bins, with all included blocks concatenated
samplePeriod : [1 x 1] Duration of each row of feat, in seconds (0.02)
blockStartStops : [nBlock x 2] Inclusive start and stop row indices in feat for each block listed in blks
events : [nTrial x 1] Row indices in feat corresponding to the Go cue/start of the tested movement epoch
presEvents : [nTrial x 1] Row indices in feat corresponding to initial stimulus presentation
reachedTargetEvents : [nTrial x 1] Row indices in feat corresponding to completion of the outward effector movement/arrival at the target posture
endHoldEvents : [nTrial x 1] Row indices in feat corresponding to the end of the target-posture hold (Stop cue)
labels : [nTrial x 4] or [nTrial x 5] Numeric trial metadata; standard Jittered Dots files have five columns and the other tasks have four, as described below
taskType : [char] Task identifier: 'JD', 'NaiveJD', or 'VE'
gripCodes : [1 x nGrip] Numeric grip codes present in the session
gripNames : {1 x nGrip} Grip names corresponding element-by-element to gripCodes
scenarioCodes : [1 x nScenario] Numeric Active/Passive condition codes present in the session
scenarioNames : {1 x nScenario} Condition names corresponding element-by-element to scenarioCodes
effectorCodes : [1 x nEffector] Numeric effector/stimulus codes, ordered from most to least anthropomorphic
effectorNames : {1 x nEffector} Effector/stimulus names corresponding element-by-element to effectorCodes

Standard Jittered Dots files additionally contain:

ratingCodes : [1 x 4] Standard ARS rating codes ([1 2 3 4])
ratingDescriptions : {4 x 1} Descriptions corresponding element-by-element to ratingCodes

labels

Rows of labels correspond one-to-one with events, presEvents, reachedTargetEvents, and endHoldEvents. The first four columns are present in every file:

1. Block number: the acquisition block containing the trial; values correspond to entries in blks.
2. Grip code: use gripCodes and gripNames for the code-to-name mapping. Across the dataset, 1 = Power, 2 = Pinch, and 4 = Relax. Code 4 is used for Empty/relax trials in the Naive Jittered Dots and Virtual Effector files. Standard Jittered Dots files contain only codes 1 and 2, including for Empty trials.
3. Scenario code: use scenarioCodes and scenarioNames for the code-to-name mapping. Across the dataset, 1 = Active and 2 = Passive.
4. Effector/stimulus code: use effectorCodes and effectorNames for the task-specific code-to-name mapping.

Only the standard Jittered Dots files contain a fifth column:

1. ARS rating: the participant's recorded Anthropomorphicity Rating Scale response. Use ratingCodes and ratingDescriptions for the four standard response categories defined above. Two trials in T11's first standard Jittered Dots session contain intermediate values (1.5 and 2.5) as recorded in labels; these are not included in ratingCodes. Naive Jittered Dots and Virtual Effector files do not have a fifth labels column and did not collect trial-level ARS ratings.

spikeTrains

spikeTrains.spikeTrains : {1 x nUnit} Each cell contains the sorted spike timestamps, in seconds, for one included unit over the concatenated session timeline
spikeTrains.slcNspTimeStamps : [nTimestamp x 1] Sample-level timing values aligning the session-level controller/task clock to the NSP clock
spikeTrains.slcNspTimeStampsPerBlock : {nAcquiredBlock x 1} The same controller-to-neural timing values separated by acquisition block
spikeTrains.nspStartTimeStamps : [nAcquiredBlock x 1] NSP start timestamp for each acquired block, in seconds

wherein:

nStep = total number of 20 ms bins in the concatenated session
nUnit = number of sorted units retained for the session/area
nTrial = number of retained trials in the session structure
nBlock = number of blocks represented in blks
nEffector = number of effector/stimulus conditions for the task
nAcquiredBlock = number of acquisition blocks represented in the neural timing structure; this can exceed nBlock when reference, practice, interrupted, or excluded blocks were acquired
nGrip = number of grip codes present in the session
nScenario = number of Active/Passive condition codes present in the session

All event and block-boundary values are MATLAB 1-based row indices into feat. To convert an index difference to elapsed time, multiply by samplePeriod (0.02 s). The files use the MATLAB Level 5 MAT-file format (MATLAB -v7) and can be read with common MAT-file libraries in Python, R, Julia, and other languages.

Code/software

The analyses described in the manuscript were performed in MATLAB R2021b. See the manuscript and SI Appendix associated with https://doi.org/10.1073/pnas.2537457123 for the analysis procedures and statistical methods. The code for reproducing the manuscript figures is made available at https://github.com/jgusman/anthropomorphicity.

Human subjects data

We confirm that the informed consent form signed by participants permits sharing of coded data that does not contain other identifying information. The included files contain no personally identifiable information (PII).

Participant identifiers are de-identified study identifiers. Sessions are reported as days since implantation ("Trial Day X"). Neural spiking data have been converted from datetime format to seconds relative to the beginning of each session.