The Channel You Cannot Mask

Many of us have done this. There is a gathering somewhere in the house, and there is a quieter place in it, and the dogs are usually in the quieter place. So that is where we are.

It gets described from the outside as avoidance, and from the inside it does not feel like avoidance at all. It feels like going where the work stops.

The explanation usually offered for this is that dogs are more empathetic than people. We went looking for that and could not get it to hold up. What we found instead was better, and it is not a compliment to anybody — it is a fact about which channel is being read.

Whatever else a gathering is, it is a place where a few specific outputs are continuously assessed. Not all of you. A band.

Face. Voice. Timing. Gaze. Posture. The rate at which you return a serve. Whether your expression is legible from across a room, and whether it is legible as the thing being felt. That band is what masking works on, and the work is real: it is authored output, produced to specification, in real time, while also having a conversation.

Not re-arguing the cost. What masking takes out of us is No. 39's subject, and it is settled there rather than restated here — including the uncomfortable half, that it cost us enormously and got some of us here alive. This spread needs only one thing from it: masking operates on the column headed assessed in a room.

In 2022, four dogs at Queen’s University Belfast were put to a question with the easy answers designed out of it: can you tell, by smell alone, whether a person was stressed when this sample was taken?

The stressor was deliberately dull. Participants counted backwards aloud from 9000 in units of 17, for three minutes, with no aids. Breath and sweat were collected before the task and again after it. Stress was confirmed independently — self-report and physiological measures — and of 53 people recruited, eleven were excluded for not having a sufficient stress response at all. The dogs were never shown a face.

Individual accuracy ran from 90.00% to 96.88%. The last of those numbers is the one to sit with: on the very first trial with a stranger's odour — nobody the dog had met, no history, no relationship, no owner, no love — the dogs were right 34 times out of 36.

The design matters more than the percentage, because it is what rules out the easy explanations. It was double-blind: nobody present knew which port held which sample, so nobody in the room could leak the answer.

Phase one trained the dog on the stress sample against two blanks. Phase two is the real question — the stress sample, the same person's baseline sample, and a blank. Not can you find a human smell. Can you tell these two states of one person apart.

Here is the sentence the authors actually wrote, and notice what it is about.

Not dogs understand our feelings. Not dogs love us enough to notice. Chemistry. A body under acute stress breathes faster, the pulse rises, the blood pressure lifts — and the mixture of compounds coming off it changes. The authors cite that mechanism as a hypothesis from earlier work rather than something they measured, and we are reporting it the same way: the profile change is inferred from the dogs' behaviour, not chemically identified in this study.

What is not in doubt is the discrimination itself. Something in the sample was different, reliably, across 36 strangers.

And a volatile organic compound is not a performance. It is metabolic exhaust. You do not choose it, rehearse it, time it, or get better at it. There is no version of you that is doing it well.

Put the two halves together and the quiet room stops being a mystery.

A room reads the authored channel and grades it. The dog reads a channel you were never able to author in the first place — and therefore cannot be marked down on.

This reframes something worth reframing. The experience of being hard to read is usually filed under deficit: something missing in the signal, something faulty in the sender. But the Belfast dogs read 36 strangers at 93.75% across 720 trials with the authored channel entirely absent — no face, no voice, no timing, no gaze.

The signal was never missing. It was being looked for in the wrong band.

That is a claim about rooms, not about noses. A room could in principle stop grading the narrow band; it mostly does not. The dog's advantage is that it never had access to the band we spend ourselves on.

If we stopped at spread seven this would be a nice zine and a slightly dishonest one. Three things have to go on the page.

One. The sample is four dogs. The authors say so themselves, in their own limitations: “It must be noted that our sample consists of only four dogs.” They add that this is in line with other bio-detection studies, because training a specialist dog takes a long time. They call the work proof of principle rather than a generalisable finding, and we are calling it that too.

Two. Discrimination is not interpretation. The study shows the dogs could tell two samples apart. It does not show they knew what the difference meant, that they connected it to the person, or that they cared. Three separate claims, and only the first one was measured.

Three. The four dogs were trained to alert. They were rewarded for indicating the stress port. That is a detection task, not a spontaneous act of concern — and reading it as concern is exactly the move this collection exists to refuse.

What we did not use, and why. The famous study in this area is the oxytocin gaze loop (Nagasawa et al., Science, 2015), and it is not on any spread here. Published commentaries flag confounds between its dog and wolf conditions and note that not all dogs showed the effect. It is contested, not refuted — and a contested result is not a foundation.

Two years later a different group asked the question we ought to be most afraid of: what does smelling our stress do to the dog?

Eighteen dogs. Odour samples from three unfamiliar volunteers, collected during stressful and during relaxing activities. Then a cognitive bias task — a food bowl at a trained rewarded location, a trained unrewarded location, and three ambiguous spots between them.

Under stress odour, the dogs became significantly less likely to approach the near-negative bowl than with no odour at all. The authors read this as possible risk-reduction behaviour. They also found the odour appeared to affect learning.

So the dog in the quiet room with us may not be being soothing. It may be being careful. That is the honest shape of it, and it is the part that stops this being a comfortable story about how much dogs love us.

In 2019 an Autistic professional dog trainer answered a question that gets asked as an insult and turns out to be a diagnosis.

The question is whether applied behaviour analysis is dog training for children. Carol Millman — B.Sc. in psychology, animal health technology diploma, a decade in practice, apprenticed at an assistance dog society — took it seriously and answered as a professional, in NeuroClastic.

Her answer is not the one the question expects. It is that the comparison is unfair to the dog trainers.

Her first task, she writes, is to make sure every one of the dog's fundamental needs is met — before any conditioning happens at all. Relieve the fear rather than train the animal not to display it. Let the retriever retrieve. Buy the terrier a sandbox instead of asking it to stop being a terrier.

There is one line in Millman's piece that does more work than the rest of it together, and it is a piece of ordinary trade knowledge.

A growl is information. Suppress the display and you have not removed the fear; you have removed the warning, and the bite arrives with no notice. Trainers have a name for this outcome because they have seen it.

Set that beside the channel argument and it clicks into place. Extinguishing an authored signal does not touch the state underneath it. The state keeps outgassing. The dogs in Belfast could smell it through the sample jar.

So here is the inversion, and it is Millman's, not ours: the professional standard for a dog is that you accept what it is, meet its needs first, and never train away a signal you need. That standard is higher than the one applied behaviour analysis has held for us.

Care with this comparison. The dog-training analogy is an indictment of ABA, never a description of Autistic people as animals to be handled. It has force only because it is made by a dog trainer, about her own trade's ethics, and because she is Autistic and writing from inside our community. Both halves matter; separate them and it becomes the insult it is meant to answer.