Verbal fluency test scored for clusters and switches

One gives six pets, six birds and six fish. The other gives eighteen unrelated animals one at a time. The count is the same and the minutes were not. Memory for this kind of material sits in pockets — farm animals, sea creatures, words sharing their first sounds — and a long list needs both a deep sweep inside a pocket and a readiness to leave it when it runs dry. Troyer, Moscovitch and Freedman split the two by scoring runs of related words as clusters and the moves between them as switches, and that is the analysis this page runs: mean cluster size describes the sweeping, the switch count describes the leaving.

On the animal minute, two neighboring words join a run when they share at least one of 25 subcategory tags — habitat, farm or pet use, family — across a list of 431 animals. A word standing alone is a run of size zero and still counts toward the mean; implementations that average only over runs of two or more report a larger mean from identical input, which is why the convention is stated. On the letter minutes three of the four published sound rules apply — same first two letters, same last three, same consonants with different vowels. The fourth joins homophones and needs a pronunciation dictionary this page does not have, so pare followed by pear reads here as a switch. Every run is drawn on screen, so you can see where each boundary went.

• Removed automatically: an exact repeat; a word that an earlier word from the same trial plus one of ten listed endings spells, so walk then walking counts once; and on a letter trial, anything not starting with that letter.
• Left to you: proper names and numbers, which the published procedure excludes and no program can spot — France and Fred are ordinary strings beginning with F.
• Shown, not deleted: an animal missing from the list goes in its own group, because every list of animals is arbitrary at the edges and absence from ours is not evidence of anything. It cannot join a cluster, since clusters run on the tags.

Every word on the results screen is a button that overrules its verdict in either direction, and the total, clusters and switches recompute from what is left. That matters because the ending rule works on spelling: give moth and then mother and the second is flagged, because in spelling it is the first plus an ending.

Healthy adults name ≈ 20 words in the animal minute and ≈ 40 words across the three letter minutes, per Tombaugh, Kozak & Rees (1999), Normative data stratified by age and education for two measures of verbal fluency: FAS and animal naming, Archives of Clinical Neuropsychology, and the same study found letter fluency leaning on education more than age while animal naming leans on age more than education — the reason a full administration runs both. Neither figure becomes a rank here. The FAS and animal-naming norms are spoken and scored live by an examiner. A typed browser version is bounded by typing speed, which is a different ability from word retrieval, and the stratified cell values could not be verified here to the digit. Compare a typed run with your own typed run a week later and the comparison holds.

Retrieval under a category is one face of the language system. The sentence completion test also asks for the right word, but narrows the search by grammar and register instead of a letter, so the two together separate a retrieval problem from a which-word-fits one. If English is not your first language, take the English level test first: a minute of retrieval in a second language measures the language long before it measures the retrieval.

This is a measurement exercise, not a clinical assessment. It reports what you did on this page against a stated reference and nothing more — it cannot establish dementia, aphasia, ADHD or any frontal-lobe problem. Only a qualified professional, working with more than a browser, can make that judgment.