Like Attracts Like
The beauty of this explanation is twofold. First, it accounts for the complex organization of the cerebral cortex (the most recent evolutionary component of the brain) using a very simple rule. Second, it deals with scaling issues very well, and indeed it also accounts for a specific phenomenon in a widespread human behavior, imitation. It explains how neurons packed themselves in the cerebral cortex and how humans relate to each other. Not a small feat.
Let's start from the brain. The idea that neurons with similar properties cluster together is theoretically appealing, because it minimizes costs associated with transmission of information. This idea is also supported by empirical evidence (it does not always happen that a theoretically appealing idea is supported by empirical data, sadly). Indeed, more than a century of a variety of brain mapping techniques demonstrated the existence of 'visual cortex' (here we find neurons that respond to visual stimuli), 'auditory cortex' (here we find neurons that respond to sounds), 'somatosensory cortex' (here we find neurons that respond to touch), and so forth. When we zoom in and look in detail at each type of cortex, we also find that the 'like attracts like' principle works well. The brain forms topographic maps. For instance, let's look at the 'motor cortex' (here we find neurons that send signals to our muscles so that we can move our body, walk, grasp things, move the eyes and explore the space surrounding us, speak, and obviously type on a keyboard, as I am doing now). In the motor cortex there is a map of the body, with neurons sending signals to hand muscles clustering together and being separate from neurons sending signals to feet or face muscles. So far, so good.
In the motor cortex, however, we also find multiple maps for the same body part (for instance, the hand). Furthermore, these multiple maps are not adjacent. What is going here? It turns out that body parts are only one of the variables that are mapped by the motor cortex. Other important variables are, for instance, different types of coordinated actions and the space sector in which the action ends. The coordinated actions that are mapped by the motor cortex belong to a number of categories, most notably defensive actions (that is, actions to defend one's own body) hand to mouth actions (important to eat and drink!), manipulative actions (using skilled finger movements to manipulate objects). The problem here is that there are multiple dimensions that are mapped onto a two-dimensional entity (we can flatten the cerebral cortex and visualize it as a surface area). This problem needs to be solved with a process of dimensionality reduction. Computational studies have shown that algorithms that do dimensionality reduction while optimizing the similarity of neighboring points (our 'like attracts like' principle) produce maps that reproduce well the complex, somewhat fractured maps described by empirical studies of the motor cortex. Thus, the principle of 'like attracts like' seems working well even when multiple dimensions must be mapped onto a two-dimensional entity (our cerebral cortex).
Let's move to human behavior. Imitation in humans is widespread and often automatic. It is important for learning and transmission of culture. We tend to align our movements (and even words!) during social interactions without even realizing it. However, we don't imitate other people in an equal way. Perhaps not surprisingly, we tend to imitate more people that are like us. Soon after birth, infants prefer faces of their own race and respond more receptively to strangers of their own race. Adults make education and even career choices that are influenced by models of their own race. This is a phenomenon called self similarity bias. Since imitation increases liking, the self similarity bias most likely influences our social preferences too. We tend to imitate others that are like us, and by doing that, we tend to like those people even more. From neurons to people, the very simple principle of 'like attracts like' has a remarkable explanatory power. This is what an elegant scientific explanation is supposed to do. To explain a lot in a simple way.