Interpersonal touch is an important channel for social and emotional interactions. We intuitively sense whether a person who touches us wants to express love or gratitude, whether they want to calm us down or get our attention. But how does our brain know what stroking, tapping or pressing means? A study now shows that two types of nerve cells in the skin probably provide the signals that are important for interpretation. The findings could also help develop prostheses “with feeling”.
Touch works like a kind of language of its own. We usually understand intuitively what feelings a person who touches us wants to convey to us. If our partner gently strokes our arm with their fingertips, we perceive it as an expression of love. On the other hand, if he puts his whole hand on our arm with light pressure, we probably have to be prepared for sad news. Previous studies have already shown that such signals are understood even by strangers – even if you cannot see the other person’s facial expression and the touch on the forearm is the only source of information.
But how does our nervous system process these finely tuned signals? What information from the nerve cells in the skin does our brain need to correctly interpret social touch?
Two types of nerve cells are particularly important
To answer these questions, a team led by Shan Xu from the University of Virginia in Charlottesville measured how the neurons in the skin react to different types of touch. In the experiment, a member of the research team touched subjects’ forearms to convey specific emotions and intentions – such as a slow stroking to reassure or a tap to get attention. The researchers used fixed touch patterns that, according to previous studies, are understood as an expression of happiness, love, gratitude or sadness or which are used to calm or attract attention. Meanwhile, the researchers measured the activity of six different types of neurons in the test subjects’ skin using fine electrodes.
Xu and her team then trained a machine learning model to map the nerve signals to different touches, similar to how our brains do. This showed that although all six types of neurons send signals to the brain, not all signals are equally important. “We found that only two types of neurons provided all the signals needed to distinguish between different types of touch,” reports Xu’s colleague Sarah McIntyre. “This suggests that these two types of neurons have a special function that helps us understand the difference between different types of social touch.”
One type of nerve cell is a so-called rapidly adapting hair follicle afferent (HFA). These sensory cells are connected to the hair follicles in the skin and react quickly to changes, such as when a hair is moved. If the hair stays in the new position, the receptors no longer send signals. The other type of nerve cells, called slowly adapting type 2 receptors (SA-II), react primarily to stretching stimuli. These receptors also signal static touch, for example from a hand held still on the skin. Together, these two types of neurons allow us to distinguish between different touches. In order to identify a touch, the machine learning model usually needed the nerve cells to fire for around three to four seconds, as the team reports.
Basis for “feeling” prostheses
“Our research is about understanding the fundamentals of how our nervous system works,” explains McIntyre. But the knowledge could also have practical applications in medicine: “If we understand how the brain interprets the signals from different types of touch, this knowledge can be helpful in the long term when it comes to correcting sensory disorders caused by age or illness. This knowledge may also be useful in the development of prosthetic limbs that interact with the nervous system and can be used by people who have lost a body part.”
How exactly the machine learning model differentiated between different touches based on nerve signals is still unclear. In future studies, the researchers want to find out exactly which mechanical properties of social touch the different types of neurons respond to.
Source: Shan Xu (University of Virginia) et al., IEEE Transactions on Affective Computing, doi: 10.1109/TAFFC.2024.3435060