I guess the most complex model is not necessarily always going to be the best one, right, as well, right? Is that right? No,

no, no, I agree with you.

There's that whole thing, the bar at all from 2013 that you've got to have a...

You've got to do the maximal model for power issues.

And then there's the other, I'm going to say the name, Matuszczak et al from 2017 that advocate for a more parsimonious model and like you build your model.

I'm kind of more inclined towards the parsimonious account.

And I think your random effects are, particularly your slopes, you want them to be theoretically justifiable.

word frequency not being yeah dispersion yeah more than frequency yeah yeah