OK but surely a traditional sentence-level or even word-level dictation would be more effective? How can you be sure that discrepancies are caused by phonological miscoding rather than by higher level flawed inferencing? For example...

— Scott Thornbury (@thornburyscott) June 7, 2022

Ah, but isn't this at the sentence level?

— Joel Swagman (@JoelswagmanJoel) June 7, 2022

For what it's worth, I regret this. I shouldn't nitpick Scott Thornbury. After all, I've learned so much from him over the years--see HERE, HERE, HERE, HERE, HERE and HERE.