Author here. Speech BCI systems are increasingly difficult to compare because they use different vocabularies, datasets, recording modalities, and metrics. We introduce a way to evaluate how much lexical information a system conveys relative to an explicit distribution over what a user may want to communicate. Happy to answer questions about the metric, the cross-study comparison, or the assumptions behind it.