Corpus-Based Transitivity Biases in Individuals with Aphasia.

The analysis focused on only 22 verbs, 11 transitively biased and 11 intransitively biased, so the results may not apply to all verbs or to other sentence structures. The speech samples came from structured AphasiaBank tasks, including narrative, picture description and procedural description, rather than from everyday conversation, so task demands may have influenced which verbs and structures appeared. People with more severe language deficits may be underrepresented because participants needed to produce speech for the AphasiaBank tasks. Ungrammatical speech was coded at surface level rather than by intended meaning, which could have affected error rates; the authors reviewed this and found that only 9 of 356 errors (2.5%) were unambiguous function-word omissions that would change transitivity. The study did not test an intervention or patient outcome, so it cannot show that using verb bias information improves communication. Some verb subsets had missing productions, such as participants with Wernicke's aphasia not producing determine, disappear or think.

Do not read the increased error rate for intransitively biased verbs in transitive structures as proof that this sentence frame caused errors for a particular patient. The study reports group-level proportions from existing speech samples and did not test treatment.