Let's suppose we were a bunch of alien scientists studying humans. We might chose to measure a whole bunch of stuff for each human, and when we do that (plotting the points in a nice multidimensional space, one point per person), we might discover that the points tend to look like two clouds. If we want to try and assign points to cloud1 or cloud2, we might find that some fancy combination of the things we have measured enables us to divide people up into 2 clouds. But that for any single thing we have measured, there is always quite a bit of overlap between the clouds.
Until one of the smarter scientists who was out to lunch says -- look at the way little humans are produced. Then we would suddenly find that things have become immensely simpler. We have essentially a binary classification, with zero overlap. And you don't have to measure a whole bunch of stuff to decide what type of human you have. Okay, says the first bunch who have spent a lot of time doing pretty pictures based on their huge multidimensional space and as many things as they could think of measuring, but there is a lot of correlation between your binary little-human-maker classification and what we have measured. Yes, says the scientist just back from lunch, but that changes over time and varies between different groups of humans. It seems to be a socially-agreed signalling mechanism about their little-human-maker status (useful since they keep covering up their little-human-making apparatus, because of the rain and wind or likelihood of mechanical damage I presume!), but like other socially-agreed mechanisms, can be subverted. Don't confuse the substance with the signal!
She then goes off to continue reading Mumsnet, which she has recently discovered -- compared to the gichy-michy on many other human modes of communication, its a breath of sanity. Although she is still wondering what a garibaldi tastes like...........