How Tabula Attentiva scores a screening session

Questionnaire items are drawn per session from banks of psychometrically matched phrasings of each construct. Selection is deterministic for a given session — an interrupted session resumes with the identical form — but differs across sessions, so no fixed, memorizable script exists. Constructs, response scales, and scoring weights are invariant across phrasings; only surface wording rotates. The exact wording served in each session is preserved in an audit trail, and adapted phrasings are footnoted on the provider report until per-variant equivalence studies are complete.

Every session runs a multivariate set of validity indicators adapted from clinical and forensic neuropsychology: performance-validity signals on the objective tasks (for example, accuracy floors on trivially easy items and response patterns faster than perception allows) and symptom-validity signals on the questionnaires (for example, endorsement of improbable symptoms and internal response consistency). Indicators aggregate conservatively: a single flag prompts caution, multiple independent flags mark the session uninterpretable. A validity flag is never treated as evidence of intent — fatigue, anxiety, literacy, and misunderstanding produce the same patterns.

Objective attention markers come from millisecond-resolution reaction times on brief go/no-go and search tasks. The distribution is summarized with ex-Gaussian components; the exponential tail (tau) captures attentional lapses, which — together with overall reaction-time variability — is the most replicated objective correlate of ADHD. Average speed alone is deliberately not the signal.

Scores are normed to the person's exact age rather than to whole school years. This directly counteracts the relative-age effect, in which the youngest children in a class are referred and diagnosed more often than the oldest for developmentally normal behavior.

Structured developmental-history items each carry an evidence-informed likelihood ratio, combined additively on the log-odds scale and capped so history can inform but never overwhelm the objective data. Deliberately neutral items are included to prevent common reasoning errors — for example, a positive response to stimulant medication is not counted as evidence, because stimulants improve attention in people without ADHD too.

Across visits, changes in the objective markers are evaluated with reliable-change indices that account for measurement error and practice effects, supporting medication-response and longitudinal monitoring rather than one-off snapshots.

Item text ships only in locales that have passed translation verification; unverified locales fall back to the reference wording. Provider reports are deterministic functions of the computed numbers — every sentence traces to an audited fact, and any optional language-model polish is constrained to rephrasing those fixed facts, never adding claims.

Tabula Attentiva is investigational and not FDA cleared. Current norms and thresholds are placeholders pending: a norming study, per-variant equivalence analyses (response-distribution and differential-item-functioning checks) for the parallel item banks, calibration of the validity indicators on honest and instructed-simulation samples, and test-retest reliability estimation for the reliable-change indices. Every report states this status.