NG7 claimed that all knowledge is held in the mind as propositions, each linking three concepts. I can offer no supporting neuroscience. Instead this piece presents some anecdotal evidence from my old-time software work which suggests at least that the position is defensible logically.
Years ago I worked on a system to capture as much information as possible about people suspected of being naughty. Part of the challenge was the variety of people, organisations, activities, financing, assets, locations, movements, communications, dates etc, and the variety of ways these could be interrelated. Pretty soon we realised that a database modelling all this in explicit detail would be complex and need lots of programming, and then would be unable to quickly accommodate new factors not revealed in the initial analysis.
The answer was to build the simplest possible database model: two entity-types, THING and LINK, with each instance of LINK joining two instances of THING. Anything could be linked to anything. Then it was possible to follow longer paths through the data to collate evidence for legal purposes. Probably a re-invention of the wheel (common in those days) but a happy client nonetheless.
Searches could be general or, using class parameters in either or both types of entity, more focussed and faster. Naturally each THING entity had content reflecting its class. Classifying each LINK entity meant that it also had some conceptual content. This made possible, for example, JOHN / LIVES / ADDRESS XXX and JOHN / WORKS / ADDRESS YYY: three concepts forming a proposition – in principal the same as we’ve already discussed for language knowledge and conscious knowledge.
The THING / LINK model seemed capable of holding all data.
Later I confirmed that by building software which translated into this form from any other database and vice versa. Why? This other client’s several recent mergers and acquisitions had accumulated lots of apparently incompatible systems. Even with no more growth it was going to take years to instal standard systems across the whole group. But the business needed to get consolidated data fast for effective control.
This was done at low cost by translating data into THING / LINK and adapting the crucial bits of software to feed off that. In addition, member companies could be given wider visibility of group activities by disseminating data – most importantly, about customers – in the other direction.
‘Data’ and ‘knowledge’ are nearly synonymous. Both words denote information – one in a computer, the other in the mind. We can be certain that a network of three-concept propositions works for data. There is a strong inference that triangles as in LS7 and LS8 work for conscious knowledge and language knowledge in the mind.
This is the most likely way that evolution allowed the storing of knowledge from endlessly varied experience. Anything more complex would need a separate program to manage it – something ruled out for the reasons given in LS5. (Computers often have more complex database models. But of course they do have stored programs to manage them.)
That’s the main message. As the piece is rather short, let’s include a little postscript.
You may have noticed that ‘network’ has occurred once or twice. It’s important and has several aspects needing to be explored carefully later in LanguidSlog, starting next time.
For now I just want to give the emerging language theory the name NG (yes, another ~G spot) from Network Grammar. There is already a blog called www.networkgrammar.com. It’s about IT stuff, not natural language, but perhaps I’m trespassing. Anyway, I’ll stick to the abbreviation and you can please yourself – ‘new grammar’, ‘nearly generative’, ‘no good’ or even my nom de guerre.