Methodology
The First Native Scale is a screening instrument based on structured parental observation, with optional age-adjusted tasks for the child. This page describes its origins, structure, scoring, and limits.
The stage model emerged from qualitative analysis of 49 in-depth diagnostic transcripts — recorded sessions with families raising children in the US, Germany, Canada, Israel, the UK, Spain, France, Norway and other countries. Across very different families, the trajectory of attrition repeated: the same behaviours appeared in the same order. The five stages are a formalization of that recurring order.
The underlying clinical archive of the initiative's partner network spans 71,838 diagnostic sessions with roughly 27,500 families; aggregate analyses of this archive are in preparation (see Data & Research).
The instrument's primary evidence is what the parent observes daily — not a one-off performance by the child. Parental questionnaires are an established, validated approach to assessing heritage-language proficiency, particularly where direct assessment in the heritage language is impractical:
Following this literature, the instrument measures exposure (how much of the language is actually present in the child's life) separately from proficiency behaviours — low exposure is reported as a risk factor, not as a "worse stage".
| Block | Items | Role |
|---|---|---|
| Context | 3 — age, years abroad, schooling language | selects age norms and the child task; no stage score |
| Exposure | 3 — language at home, contacts beyond parents, media | risk factor, reported separately; no stage score |
| Observation | 6 — language of reply, talk about feelings, retrieval pauses, attitude, conversation with grandparents, comprehension | determines the stage |
| Child task (optional) | 12 picture-naming items (ages 3–9) or 8 verbal tasks (ages 10–13) | refines vocabulary and fluency; never determines the stage alone |
Observation items are phrased as behaviour, not evaluation — "When you address the child in your language, in which language do they reply?" rather than "How good is their language?". Each observation item is scored 0–3; the stage is derived from the proportion of the maximum score, so that unanswerable items (e.g. no living grandparents) are excluded without penalty.
A child of 4 and a child of 12 lose different things. At four, object vocabulary goes; at twelve, object vocabulary usually survives while abstract vocabulary, figurative meaning and connected speech collapse. The task therefore differs by age band:
| Ages | Task | Expected range |
|---|---|---|
| 3–5 | 12 pictures, basic objects | 9–12 named |
| 6–9 | 12 pictures incl. lower-frequency household objects | 10–12 named |
| 10–13 | 8 verbal tasks: explaining abstract words, idioms, antonyms, connected narrative | rated per task; switching to the majority language is the most informative signal |
The result screen shows not only the stage but the specific parental answers it follows from ("you noted that the child replies in the majority language; that talking about feelings switches languages; …"). A screening tool that cannot show its reasoning invites justified distrust.
Current stage thresholds are expert-set, informed by the qualitative core of 49 transcripts. Ongoing calibration uses anonymized questionnaire results (answer indices only — no names, no contact data, no free text) compared against subsequent specialist assessment, with thresholds to be adjusted as the calibration sample grows. This status is stated openly wherever the instrument is offered.
First Native