Test a correspondence instead of trusting resemblance
Before you begin: Sounds & pronunciation. The small dataset below is invented for practice, not evidence about a real language family.
Historical comparison looks for recurring correspondences in comparable words, then checks alternatives such as borrowing or chance. One similar-looking word cannot establish shared ancestry. A proposed sound relationship should make predictions about additional data.
Variation within a language also follows patterns. A dialect is not a defective version of a standard. Keep the speaker, place, context and date of evidence visible; do not infer a whole community’s grammar from one quotation.
Worked example & observations
Original synthetic dataset with matched meanings: variety A has pala ‘leaf’, pino ‘stone’, tami ‘path’; variety B has fala ‘leaf’, fino ‘stone’, tami ‘path’. The first two pairs suggest initial p in A corresponding to f in B, while the third shows that not every initial consonant changes. This is a candidate correspondence, not a reconstructed family tree.
Your turn
Predict B’s form corresponding to A’s paku ‘basket’ under that hypothesis. If new evidence gives B paku, what should you do?
Show answer and reasoning
The simple prediction is faku. An observed paku is counterevidence to the unrestricted rule. Check the transcription, loanword history and conditioning environment, then revise or reject the rule. Do not silently discard a pair because it disagrees.
Watch for: Matching spelling is weaker than matching analysed sounds, and matching sounds alone is still insufficient. These four invented pairs cannot establish ancestry or a direction of change.
Published and planned
Published: this starter lesson, worked example and exercise with an answer.
Planned: Documented sound changes, borrowing, dated attestations and ethically collected variation data.
Validation: source review and mathematical checks where applicable. Photography captures, audio listening and device workflows have not been performed for these lessons. Synthetic cases are labelled; no personal media or field observations are implied.
References
Original AUWEN teaching and examples. Sources reviewed 14 September 2026.
Learning update log →