About this database
Language Elements is a living systematic review database of neurostimulation studies examining the causal role of brain regions in language processing. It compiles findings from transcranial magnetic stimulation (TMS), transcranial electrical stimulation (tES), and direct electrical stimulation (DES) studies — the only techniques capable of establishing causal brain–language relationships rather than correlational ones.
The database is built around elementalism: a theoretical framework that characterises brain regions by the minimal computational operations they causally support, inferred bottom-up across heterogeneous tasks (following Genon et al., 2018). This high-specificity approach distinguishes Language Elements from existing resources, which typically catalogue findings at the level of broad linguistic domains.
The database is designed to serve both basic research — supporting experimental planning and meta-analytic synthesis — and intraoperative language mapping, providing a searchable evidence base of tasks documented in the neurostimulation literature for each brain region.
The database
The systematic review (PROSPERO: CRD42024602006) searched PubMed, Scopus, Embase, and PsychInfo from October 2024 to January 2025, returning 12,763 records. After deduplication and screening, the current database includes:
Updated 28 April 2026
The database is live and updated as screening and extraction continue.
How to cite
A paper describing the database and the elementalism framework is currently in submission. In the meantime, please cite the PROSPERO registration:
Key references: De Witte et al. (2015) DuLIP — doi:10.1080/02687038.2015.1071993 · Wager et al. (2017) Operating Environment — doi:10.1016/j.neuchi.2016.10.002
Contributors
Language Elements is an international collaboration between research teams in the UK and Germany.
Funding
This work was supported by the British Council Going Global Partnerships Springboard Programme (UK–Germany).
Contact
For queries about the database, the systematic review, or potential collaborations, please contact T. R. Williamson at t.williamson@uwe.ac.uk.
The Database tab searches the literature directly. Every result is drawn verbatim from the systematic review dataset — no inference, no generation. Free-text search and filters (stimulation type, linguistic area, hemisphere, inhibition/facilitation) operate on the raw data. The brain visualiser plots MNI coordinates extracted directly from the included papers.
Type a query to begin. Elements mode accepts searches across six categories: brain region, task name, stimulus type, language, linguistic domain, or reported effect of stimulation. The tool interprets the query automatically and returns a functional profile of the relevant studies alongside several context summaries.
The search bar classifies the query into one of six categories and shows the resolved category below the bar, with an override dropdown if the interpretation is wrong. Region, language, and linguistic domain queries are resolved via synonym tables. Task, stimulus, and reported-effect queries are resolved via an AI inference pass that reasons about the construct behind the query — so a search for “picture naming” returns papers whose paradigms instantiate picture naming, not only those that use the string verbatim.
The query is routed to a region, language, or linguistic domain synonym table first; if a match is found, results are returned instantly with no network request. If no match is found, an AI call infers whether the query is a task, stimulus, or reported effect and caches the result for subsequent searches.
For task, stimulus, and reported-effect searches, the tool reasons about the neuropsychological construct behind the query, groups filtered papers into construct clusters, and returns per-paper rationales for inclusion. Results are cached per query.
The tool identifies the elements — minimal computational operations — that the filtered papers collectively support. Elements are inferred bottom-up from the data, not assigned top-down from a fixed ontology. Each element card shows the studies grouping under it, their combined sample size, and a confidence indicator. On each card, “What is this element?” returns a plain-language definition; “How was this element derived?” returns the reasoning chain from the grouped studies.
Elements mode uses a two-stage AI pipeline to infer element labels bottom-up from the filtered studies. Operation-function labels are generated following Genon et al. (2018) functional characterisation. Label specificity is calibrated to the neuroanatomical hierarchy level of the search. Results may vary slightly between sessions due to the probabilistic nature of language models.
If the AI call fails — for example, because of a network issue or an API outage — Elements mode falls back to a flat, ranked list of all identified processes ordered by study count. All data remains drawn directly from the systematic review.
Below the elements, five summary cards describe what the filtered papers report for the five datapoints not searched. Each summary is a short prose synthesis; each has a “Breakdown” button that opens a detailed view in the right panel. The right panel stacks — multiple breakdowns and element explanations can be open at once, each with its own close button. Region breakdowns render as a three-level neuroanatomical hierarchy (lobe → gyrus → Brodmann area) with paper counts rolled up at each level. Task, stimulus, and effect breakdowns render as construct clusters with member papers grouped under canonical construct labels.
Type a brain region name to begin. The tool runs in two stages: first summarising what the neurostimulation literature reports for the region, then surfacing tasks from the database that studies have used at or near that site.
When you search a region, the tool identifies all neurostimulation studies in the database targeting that region and synthesises the linguistic processes the studies implicate. These processes are grouped into summary element labels using a two-stage pipeline with a defined fallback:
When you search a region, Tasks mode uses a two-stage AI pipeline to summarise the linguistic processes studied at that site. Summary element labels are generated using bottom-up functional characterisation guided by Genon et al. (2018). Label specificity is calibrated to the neuroanatomical hierarchy level of the searched region. Results may vary slightly between sessions due to the probabilistic nature of language models.
If the AI call fails, Tasks mode falls back to a flat, ranked list of all identified processes ordered by study count. All data remains drawn directly from the systematic review.
Once a region has been summarised, the tool surfaces tasks from the database that studies have used at or near that region, in three steps:
All tasks used in neurostimulation studies of this region are retrieved from the database and organised by the processes they target. Tasks are deduplicated and prepared for the AI, which receives the full task evidence for this region alongside the region summary from the previous stage.
An AI model selects and ranks tasks from the database evidence, applying clinical constraints drawn from two peer-reviewed frameworks: De Witte et al. (2015) (DuLIP) and Wager et al. (2017) (operating environment). All tasks shown cite their source papers. Where the database lacks a suitable task for a specific process, the model may surface a DuLIP fallback task (De Witte et al., 2015) — clearly labelled as such — so that every process identified in the region summary has at least one candidate task, drawn either from the systematic review literature or the DuLIP protocol.
If the AI call fails, no tasks are shown. The region summary from the previous stage remains visible and can still be used to support the clinician's own literature review.
The database-derived task lists presented by this tool are intended to support the clinician's own literature review, not replace clinical judgement. They are drawn from the neurostimulation research literature and have not been prospectively validated as clinical decision-support outputs. All task selection for awake craniotomy procedures must be approved by the responsible surgical and neuropsychology team. Confidence scores are heuristic indicators based on study count, stimulation modality, and sample size — they are not validated clinical risk scores.