Birch is a professor of philosophy at the London School of Economics and author of "The Edge of Sentience: Risk and Precaution in Humans, Other Animals, and AI."

A stroke, a traumatic brain injury -- at any moment, any one of us could leave behind our normal state of consciousness and start to require 'round-the-clock care. Sometimes, a patient displays regular cycles of wakefulness and sleep without regaining the capacity for normal voluntary action. This is a tragic situation for the patient's family -- but what, if anything, might the patient themselves be experiencing?

It's impossible to say with certainty. Patients may smile, grimace, cry, grunt, or groan while awake, and families will often take these as signs of pain and pleasure. But their doctors usually remain skeptical, suspecting these behaviors are merely reflexes.

Before 2002, all these patients were described as "vegetative." Yet it became apparent to experts that some of them were still having conscious experiences, so the term "minimally conscious state" was introduced to cover these cases. That, however, just left doctors with the challenge of distinguishing "minimally conscious" from "vegetative" patients, a frankly impossible task.

Lists of criteria, such as the Coma Recovery Scale-Revised, attempt to standardize subjective assessment. But the key questions they rely upon to disentangle "conscious" from "reflex" behavior include: Can the patient fixate on an object for more than two seconds in at least two out of four trials? If the doctor applies pressure to a finger or toe, does the patient move their opposite arm or leg to the site of the pressure on at least two out of four trials? I don't think anyone imagines these tests reliably track consciousness. They are useful up a point, but they don't tell us whether the patient is having experiences.

In fact, even when a patient displays no outward signs of conscious experience at all, "covert" conscious experience may remain. In a celebrated technique developed in 2010 by neuroscientist Adrian Owen's group, patients are placed in an fMRI scanner and asked yes/no questions. They are instructed: if yes, imagine playing tennis; if no, imagine walking round your house. These tasks light up very different areas of the brain, creating an fMRI signature of the answer in healthy adults. And some of the so-called "vegetative" patients could also answer questions in this way.

This led to a flurry of interest in developing cheaper EEG-based methods for detecting covert responsiveness -- methods that can be used at the bedside in the ICU. These pioneering techniques seem to be picking up signs of responsiveness in around 25% of outwardly "unresponsive" patients, according to the most recent study.

The risk of false positives -- where we mistake statistical noise for responding -- must be taken seriously, with uncertainty communicated sensitively to the patient's family. At the same time, there remains a risk of false negatives: cases where the patient has experiences but cannot follow instructions, not even through brain activity.

Because we cannot directly access the "ground truth" of whether the patient is conscious, we cannot put reliable numbers on the false positives and false negatives. Yes, the true rate of residual conscious experience might be well below 25% -- but it could also be far above 25%.

What should this mean for clinical practice? A task force in 2010 recommended replacing the term "vegetative state" with "unresponsive wakefulness syndrome." The American Academy of Neurology now uses both terms. The new term is a step in the right direction, because it removes the stigma of "vegetative," but I don't think it goes far enough.

The fundamental problem is not the word, but the practice of writing off a group of patients as "not even minimally conscious" when we are far from certain of this. The term "unresponsive wakefulness syndrome" still has that regrettable implication, because it is still used interchangeably with "vegetative" and contrasted with "minimally conscious state."

We need to err on the side of caution in these cases, and that means phasing out -- in therapeutic, legal, and colloquial contexts -- any category that starkly implies "not even minimally conscious." We are not in a position to make that diagnosis, and we need to have humility about this.

A way forward, for now, is to use the more general category of "prolonged disorder of consciousness" and then add notes describing the specific condition and needs of each individual patient. We should also support the development, over the long term, of more nuanced, fine-grained diagnostic frameworks that can evolve together with our scientific understanding of consciousness.

A residual capacity for pain, pleasure, and sensory experience is always a realistic possibility. So, we have to take common-sense precautions to allow for that possibility. This means using pain relief whenever a procedure has the potential to cause pain, and it means explaining to the patient what is happening to them.

Sometimes a patient's family and clinical team agree that it is in the patient's own best interests to be allowed to die, based on what is known about their wishes (often from an advance directive). These decisions should not hinge on whether the patient is "vegetative" or "minimally conscious." In these discussions -- which will always be agonizing for all involved -- it should not be assumed that the patient feels nothing.

In the U.K., clinical guidelines are starting to move in the right direction, and a High Court judge recently described the term "vegetative state" as dehumanizing and offensive.

Sadly, doctors in the U.K, U.S. and elsewhere are still expected to make a diagnosis of "vegetative/UWS" or "minimally conscious." But when it comes to treatment, guidelines emphasize that risks of continuing pain and pleasure should always be taken seriously. The fact that you think a patient may be wholly unconscious is never a sufficient reason to withhold pain relief.

That's the precautionary attitude we need to cultivate with these patients -- an attitude where everyone is treated with respect, and no one is written off as "not even minimally conscious" on the basis of highly uncertain criteria.