‘Family-Resemblance concept’ is a contradiction

NOTE HXA7241 2020-09-27T12:44Z

Either you have a Family-Resemblance local network relation, or you have some items with a commonality. The one is not the other.

What is Family-Resemblance (as popularised by Wittgenstein)? FR is a network, with each member sharing nothing except pair-wise similarity/relation. The members are connected only by those links, which must each be different. Elements of a FR have nothing in common overall ‒ that is its essence.

The letter sequences example of an FR is misleading. Eg:


(note the first and last letters of each ‘word’ are the only links between the words ‒ all connected, but no part is common in all, and each link is different). This illustration only works because a commonality is smuggled in in plain sight that helps us recognise the elements as a group: they are all words, they all have a linkage template of a letter, and they are all presented together. Otherwise, how do you know it is those words you should attend to, how would you know that they are the only things to check? Why not check the other words in the text around them? And why just words? Why can it not be one of those trick questions, where it is really about the shapes of the letters? Why could not the links be anything? …


The question to ask is: how would you know whether some new object is a member or not? Examining this problem shows that for a FR the answer is intractable.

First, you will realise that you have to inspect every feature of the new candidate to compare it with every feature of every current member, to check for potential linkableness. (Remember: there is no single common feature so everything must be examined.). But no, that is not enough, because what if there is an intermediate linkage somewhere else that will join the two being compared? Well, now this is getting difficult ‒ you have to inspect and compare with every feature of every other object in the universe! But even then you might be missing something because of the transitiveness: the intermediate bridge might have to be built from multiple links. So now you have to check every combination of every object in the universe too.

This is intractable and obviously not how groupings are made. It is impossible that FR is how cognition works, or any modelling or computational system. Any FR presented as a group can only be a group for some other (perhaps non-obvious) reason. (Maybe just rote learning of an enumerated list ‒ what is common is simply that we (all) put them together (accidentally).).

FR means no commonality, which means no restriction in general on which items can link to which others, and that means the network is boundless. And we reach the conclusion that in truth there is only one FR: the whole universe. It is practically always possible to link anything to anything else: you just choose the right abstraction. At the limit, you can simply note that A is in the same universe as B, and that is what links them ‒ which applies to everything, and is not disallowed.


Are games, eg, a FR? It is a pointless question in this context. They could be an arbitrary miscellany, but they could also equally possibly have a commonality. It is entirely contingent and tells us nothing philosophically.

The better analysis, compared to ‘family resemblance’, is to say that:

  • some patterns, that we can competently grasp, are obscure and difficult to articulate;
  • and, overlappingly, we can be mistaken about whether there is a pattern in any particular place or not.

This is just forced by the consistency of the concepts themselves: a pattern means something underlying in common. We do not need to invent a new kind of group that is simultaneously not a group, and indeed we cannot.


The FR structure overall must, by definition, be made of different pieces. So it cannot be regarded as having similarity overall, and so is not a group/concept. All things being similar, all things sharing something in common, is simply a different structure to FR. If you start with a thing, then find something similar, then find a third thing that you deliberately choose for it being different to the first, of course you do not produce a group that has an overall similarity.