Tony Kroch (1946-2021) was a linguist, a free thinker, and a friend. I like to think of myself as one of Tony's "adopted academic grand-children" for one of my supervisors, Susan Pintzuk, did her PhD with him. I also could spend some time with him during an academic year at UPenn in 2014-5, which, in no small part thanks to him, I count as one of the happiest times in my life. Sadly, he recently passed away.

Beatrice Santorini, one of Tony's collaborators and friends, sent me a link to one of his talks and I took the liberty to cut the video and put it up online as a way to remember him. His talk, recorded in 2003 at a symposium to honor the legacy of Zellig Harris, is entitled Distributional Evidence, Imperfect Learning, and Language Change. You can watch the entire video here.

There is so much to say about Tony's ideas. Let's consider just one of a million possible things: Take any two competing linguistic models (Distributional Morphology vs. Finite State Transducers, Lexical Functional Grammar vs. Berkeley Construction Grammar, etc. etc.) - how can we ever decide which one is to be preferred over the other? I don't think anybody knows. Can one be "true", another "false"? How is this possible in the realm of formalisms and abstract concepts? Or are they just alternatives of each other? Can one translate between them to a large degree? Tony thought about this problem in the context of sociolinguistic variables vs. Grammar Competition for a bit and summarized his view in the very last sentence of the talk, "Whether the one or the other way is the right way to think about it is a tricky question. But I don't think they're the same way of talking." I like this. It resonates with me in its elegance and honesty. It's unpretentious but also based on deep understanding. And the things he said just might turn out to be correct, too...