Long-Term Intracortical Speech BCI (Card et al., 2026)
Measured by Card, Singer-Clark, Peracha, Stavisky, Brandman et al. · Nature Medicine (2026)
Inputs
The measured or assumed values behind the calculations, each with its source.
- N = 125000
- 125,000-word vocabulary for the brain-to-text speech decoder (Abstract; Speech decoding section)
- rate = 56 word/min
- Average rate across 183,060 personal-use sentences totaling 1,960,163 words (Abstract)
- P = 0.992
- 99.2% word accuracy reported for the transformer-based decoder in prompted word-copy benchmarking. Paired here with the personal-use average rate only as a mixed-task operating-point estimate, not as a single directly reported throughput.
Strictest ITR
Each scoring method is an upper bound on the channel, so the headline is the strictest (smallest) one for this entry. Use the score selector on the home page to view any single method across entries.
-
Error-corrected words per minute
P x rate = 0.992 x 56 = 55.6 net word/min
The 56 WPM value is the reported personal-use average; the 99.2% word accuracy comes from prompted benchmarking. This is a mixed-task estimate built from two reported operating points, not a single task's directly reported ITR.
-
Shannon per-word entropy of English
H ~= 5.0 bits/word
Credits only the information in the English produced, independent of vocabulary size.
-
Information transfer rate
55.6 word/min × 5.0 bits/word ÷ 60 s/min = 4.63 bits/s
What counts as a bit depends on the action space. The number of distinguishable actions and how likely each one is are design choices of the task, not the sensing hardware. The same modality can present a fixed set of targets, a set pruned per step by a grammar or language model, or a continuous control space. Each of these changes how many actions are live and how the probability mass is spread, and therefore the information per selection. Read the action space below before comparing headline numbers across entries.
Action space
What the user can produce at each step, and how those options are distributed.
- Structure
- Context-dependent (the live set changes per step)
- Size
- 125000 distinguishable actions
- Prior
- Context-conditioned: likelihoods depend on prior actions
- Notes
- The speech channel is a large-vocabulary brain-to-text decoder with a language model over more than 125,000 words. The live word probabilities are context-conditioned, so the 125k vocabulary is a nominal action space rather than a uniform set. The same implanted arrays also drove a 2D neural cursor; that cursor benchmark is split into its own entry, matching the channel-vs-application split used for BrainGate2.
Comparability The strictest bound here is the Shannon entropy of the output text, under one predictor held constant across the whole atlas (≈1 bit per character). That shared predictor makes it directly comparable to every other text entry (keyboards, spellers, silent speech and speech BCIs) regardless of prior or vocabulary size. For most text interfaces it comes out tighter than the raw-selection bounds, but not always. Where a small vocabulary makes Wolpaw tighter, that wins instead. Any Fitts, Wolpaw or log₂(N) figure shown below is another bound on the same channel. Switch the home-page score selector to compare one across entries.
Other score types
Bounds the atlas keeps out of the default strictest headline: as-reported figures, alternate task conditions, or raw-channel ceilings that shouldn't win the headline by default. Each still carries a score type, so the home-page selector ranks this entry on it when you choose that type. Read its derivation before comparing across entries.
-
Bits per selection (Wolpaw formula)
B = log2(N) + P*log2(P) + (1-P)*log2((1-P)/(N-1)) = log2(125000) + 0.992*log2(0.992) + 0.008*log2(0.008/124999) = 16.729 bits / selection
Term 1 is the information if every choice were correct; terms 2-3 subtract the bits lost to the error rate, assumed spread evenly over the other N-1 targets.
-
Selections per second
T = 1.07143 s/selection -> 1 / 1.07143 = 0.933 selections/s
-
Information transfer rate
ITR = B * selections/s = 16.729 * 0.933 = 15.614 bits/s
-
Achieved-bitrate credit per net-correct word
N = 125,000 → log2(N − 1) = log2(124,999) ≈ 16.93 bits per net-correct selection (Nuyujukian 2015, which introduced the metric, used log2(N); at this N the difference is negligible).
-
Net-correct word rate
net-correct = 2P − 1 = 2(0.992) − 1 = 0.984 of words. At 56 word/min (1.071 s/word) → 0.984 × 56 / 60 = 0.919 correct/s.
A word error commits the wrong word rather than timing out, so incorrect = 1 − P. Same N (125,000), word accuracy (99.2%) and rate (56 wpm) as the entry's Wolpaw calc. As that calc's own caveat notes, the 125k-word vocabulary is context-reweighted by the language model each step, so feeding it into log2(N − 1) is a large-vocabulary capacity view (~17 bits/word) — raw channel capacity, not communication, analogous to the nagel code-space figure. The ranked communication rate is the 4.63 bits/s word-entropy Shannon.
-
Achieved bitrate
16.93 bits × 0.919 correct/s = 15.6 bits/s.
Source
- Authors
- Card, Singer-Clark, Peracha, Stavisky, Brandman et al.
- Publication
- Nature Medicine, 2026
- Reference
- Nature Medicine full text