How does the brain define concepts?

I suppose it would be either by essential features, family resemblance, or function, or some combination of these.
I’m asking because I’m going to write a paper on family resemblance definitions of religion and think it would be relevant to discuss how the concept ‘religion’ is created in minds. Cites would be great!

It’s very likely that no one knows. Our understanding of the brain is primitive at best, and it’s very easy to ask questions that can’t be answered.

In terms of other concepts, presumably. In some sense, all the mind is is a very complex set of relationships between concepts.

You could try prototype theory.

I was going to suggest this. I’ve done extensive studying into computer vision and a lot of that involves trying to figure out how the human brain interprets images and then trying to emulate that process with computers. It’s been a while since I’ve read up on it, but I believe that this is still the best guess at how the brain does it and a lot of algorithms are based on it, everything from basic recognition to compression.

Basically, the way it works is that you get a representative object or set of objects then calculate the differences and based upon a threshold, you can determine if it’s in the same class or not. A simple algorithm example of this is video compression, where a representative frame is taken and nearby frames are stored as differences of that frame. When frames differ by too much, a new representative frame is stored and the process is repeated.

Face recognition is almost certainly done this way. We recognize the basic structures of a face from a very young age. Personally, since it’s something I pay a lot of attention too, I’ve actually noticed that I have certain classes of people and, when I think about it, I realize that some new person I meet reminds me either looks a lot like a certain group, or even a specific person that several other people look a lot like. Music is another good example, except we explicitly classify it into genres.

The thing is, one of the things that the human brain is very good at but computers are much less good at is classification. Again, faces are an excellent example, where even very good algorithms have a fairly high failure rate (as much as 5%, more or less depending on resolution, angles, lighting, and all that), but a human can pretty easily pick out similar faces, even with varying angles, lighting, and all that.
Venturing a bit out of GQ into a guess, I think this is probably also how more abstract concepts are stored. For instance, western culture is highly familiar with Christianity, and so the most basic principles of it are often used when comparing religions, even when comparing very different religions. At the same time, take an East Asian person, and they’re probably a lot more familiar with Buddhism and they’ll take that as a prototype and differentiate from there.

Anyway, hope that’s helpful, if a bit rambly.

For some values of “this way”. When a computer compares how similar two images are, it’ll usually just do it on a pixel-by-pixel basis, but our brains work more on an object-by-object basis. If you see a front view of a person, and then a profile, you’ll certainly recognize both as faces, and probably recognize them as being the same person, even though most of the pixels will be completely different.

I don’t want to hijack too much, but object comparison is quite a bit more complicated than a simple pixel by pixel comparison. An eye, for instance, is relatively easily recognized because it has a distinct shape, easily identified corners, has sharp edges (skin tone to white), and a well established relationship to other facial features. There are some issues with orientation, but identifying points of interest (corners of the eyes being very easy to find) particularly when surrounded by skin tone colors, provided the face can be identified, the face can then be adjusted based upon those points of interest to then be able to apply those points of interest in comparison to the template.

Yes, it’s definitely not anywhere near as advanced as the way the human mind interprets and understands objects, and it certainly won’t be able to infer similarity to other objects without extensive training and other pattern recognition techniques. I have some interesting articles about these techniques somewhere with some interesting results, having pretty good success rates at recongizing objects despite scaling, rotation, and orientation, by identifying points of interest and transforming objects into a common space for comparison.

Anyway, my underlying point was more about trying to provide an analogy about how the process works, and I hope at least the video compression one helped clarify the idea.

I should have been more clear with my “usually” statement. Yes, computers can do better than that for some object-recognition tasks, but for encoding a movie via changes from key frames, they probably aren’t doing much more than the pixel-by-pixel comparisons (though some of them do seem to track translations of pixels, to at least some degree).

Also I’d be careful of the idea of “concepts”.

<maybe off topic>
I find that many people when learning that the brain stores a lot of information by association, just jump to the conclusion that there is some sort of hierarchy. And indeed, that there are central “facts” which we must learn before we can learn anything else.

But the kinds of association we’re talking about start with just “this shape is associated with this colour”, “this noise is associated with that noise”.
When it comes to more complex knowledge, we don’t know how it works. But there are lots of reasons to suppose it does not begin with immutable axioms.

yelimS - Nobody knows how the brain defines concepts in physiological terms, nor do we know much at all about how human brains actually work. But reading your OP again I wondered if you were looking for an answer regarding human psychology, or how we define and develop ideas and concepts at a higher level. Is this what you were looking for?

That may well be, but the only short and honest answer remains that we just don’t know (and, we do not really know what the word “concept” refers to, if anything). In cognitive psychology there are several rival theories (of which prototype theory, mentioned above, is one) but no consensus as to which (if any) is right, and it is by no means clear whether the rival theories are really addressing the same issues.

Another way of putting this is that, despite the extensive application of experimental methods to the problem, it remains, essentially, a philosophical issue. (So come back in a few centuries and, with luck, we have the beginnings of an answer for you.)

Thanks for all the answers! I’m only indirectly interested in physiology. The idea is that scientific definitions of religion are often criticized for leaving out some ‘obvious’ religions, or including ones that ‘obviously’ aren’t. Yet I’ve never seen anyone suggest we go about the discrepancy the other way - by looking at how the brain constructs this concept of religion that doesn’t match the theoretical definitions. My assumption is that it does so in a family resemblance sort of way, like prototype theory, but I’ve no clue where to get cites for this.
I have other arguments for family resemblance definitions of religion, too, but I think a good way to support this, would be to show that it is the way the brain creates concepts.
So: interesting answers - does anyone have a cite of any kind? Someone with a bit of authority saying that prototype theory is a major theory in the field would probably do, as it’s not for my main line of argument.

Something we do know that the brain excels at is pattern matching and general categorization.

Without going into the specifics of how it might do it - I think it is safe to say that whatever the brain does it is messier and more complex and different than any kind of theory like “prototype theory”

If you read up on classification systems - like neural networks and support vector machines you will see that there can be classification based on training that isn’t really definable in terms of some high level theory that follows logical rules, etc.