The Channel With No Task

The skin does not run one touch system. It runs two, side by side, on the same surface, and they are built to opposite specs. The familiar one is insulated, fast and precise. The other is uninsulated, slow, and found only in hairy skin. Take the conduction speed. The fast system away from the skin carries signals at roughly 35 to 75 metres per second. The slow system runs at roughly 0.5 to 2 metres per second. That is a gap of about fifty times, and it is the whole design brief. Fifty times slower is not a worse sensor. It is a different channel, wired to a different destination, for a different purpose.

Run the arithmetic on a two-metre body. A signal at 50 metres per second crosses it in about 40 milliseconds, which is fast enough to sit inside a reflex loop and fast enough to catch a dropped tool. A signal at 1 metre per second takes about two seconds to make the same trip. Nothing useful happens on a two-second loop in manipulation. You cannot catch anything with it. You cannot steer a grip with it. The slow system was characterised in the published literature far later than the fast one, Olausson and colleagues in 2002 and Loken and colleagues in 2009 on the tuning, and it is close to absent from the engineering literature that followed. That is not an oversight in the biology. It is the shape of the thing. It was never built to act.

Where the fast system answers every mechanical modality, the slow one answers a narrow band. Published tuning runs like this. Below 0.5 centimetres per second the response is minimal. From 1 to 3 it rises. From 3 to 5 it peaks, which is the optimum. From 5 to 10 it holds sustained and high. Above 10 it declines. The peak response sits at 1 to 10 centimetres per second, and the temperature condition is as tight as the speed: skin temperature, around 32 degrees. Now look at what falls inside that band. The stroke speed of a human caress, in bonding behaviour between a parent and a child or between partners, lands inside it. The system is a detector for a specific gesture, and the gesture is care. Slower does not read as gentler. Faster does not read as firmer. Both fail.

Follow the fibre and the separation becomes structural. The fast system runs to the somatosensory cortex through the dorsal column and the medial lemniscus, and it builds a map. The slow one runs through the spinothalamic tract to different ground entirely: the posterior insula, which carries the felt state of the body and produces the sense that this touch feels good; the orbitofrontal cortex, which values the reward; the anterior cingulate, which carries social salience; and the amygdala, which learns safety. There is no map at the end of that route. The receptive field is small, around a square centimetre, and the spatial resolution is poor by design. Ask the slow channel where you were touched and it cannot answer. Ask it how hard and it cannot answer that either. Ask it whether you are safe and it answers immediately.

One system reports the object. The other reports the relationship.

Almost all instrumented skin targets the fast system, and the four metrics say so plainly: two-point discrimination, force sensitivity and dynamic range, bandwidth measured in taxel count, and durability with conformability. Every one of those is a discriminative requirement, and every one is necessary for manipulation. None of them is sufficient for contact with a person. A care robot carrying the best discriminative skin in the world can measure a pulse through a fingertip, detect the moment a grasped object starts to slip, and map the geometry of a surface it is laid against. It cannot deliver the emotional content of a caress. It cannot signal safety through gentle skin-temperature contact. It cannot adjust its own touch so that it lands as caring rather than clinical. The probability that a person can tell the difference is not zero. The field prices it at zero.

The engineering answer inverts the usual order. Compliance comes first and resolution second, because for this channel a soft warm surface with coarse pressure sensing beats a rigid high-resolution array. Temperature stops being a comfort feature and becomes a parameter: 32 to 34 degrees reads as pleasant against room temperature. Timing becomes the signal, which means the arm needs velocity profiles of 1 to 10 centimetres per second and normal force of roughly 0.5 to 2 newtons. And two modalities drop out of the toolkit. Electrotactile and vibrotactile stimulation drives the fast fibres, not this one, so neither can produce the felt sense of caring touch no matter how it is tuned. What is missing is a transducer class that has barely been designed: soft, compliant, temperature-controlled, low-bandwidth and low-force. I have built companies across twelve countries and restructured a industrial group, and the pattern holds in every system I have worked inside. What gets measured gets funded, and what has no task gets no line on the roadmap. The channel has no task. That is why nobody built it.