The Semantics of Line Drawings XIII, Why is Reading so Easy?

[ The Semantics of Line Drawings XII, Grylloi | The Semantics of Line Drawings XIV, Grylloi and Visual Linguistics ]

I'll return to grylloi, but I want to ask a question. Why is reading so easy? In a few seconds last night while updating this blog, I was able to read wikiHow's recipe for linking Blogger to Twitter. And therefore, as Mark Changizi puts it in his book The Vision Revolution, install new software in me. Amazon, he says, listed at least 720,000 "how-to" books when he was writing in 2009. Each is a piece of software just waiting to be implemented by a human brain. Recipes for gefilte fish; techniques for training your labrador to fetch; how to clean the dirt off the bottom of a swimming pool. All encoded as a sequence of little marks that impress themselves on our brain, automatically and without effort.

But what is it about letters and words that fits our visual system so well? Changizi's answer is that the geometrical properties of written words are similar to those of objects. Our brains have had to evolve to be very very good at recognising objects; if the brain can use the same part of the visual system for recognising words, it's bound to help it read them.

By "geometrical properties", Changizi means how often different kinds of junction occur in words. This, obviously, is a consequence of the kinds of junction in the symbols that make up the words. These symbols might be letters such as A or ҕ, syllabic characters such as those in ᓀᐦᐃᔭᐍᐏᐣ (the Plains Cree word for Plains Cree), logograms such as the Chinese 中文, and so on. A comprehensive list of writing systems can be found at Omniglot's "Index of languages by writing system". Look at these, and you'll see that, for example, Y junctions are common but ✱ junctions are rare. Somehow, all these symbols feel as though they come from the same family of shapes.

Why? Our visual system is organised in a hierarchy. At the top is recognition of complete objects. At the bottom is recognition of primitive features such as single edges or strokes. And in between are levels that recognise simple combinations of these primitives, such as the L, T, and Y junctions I've marked in my cartoon below:

"A4Billion" by me

In fact, most of the junctions in my drawing are L, T, and Y junctions. There are good geometrical reasons for this, to do with how objects meet and overlap, and how the places where they do look when projected down to two dimensions. This is explained in The Structures of Letters and Symbols throughout Human History Are Selected to Match Those Found in Objects in Natural Scenes by Mark Changizi, Qiong Zhang, Hao Ye, and Shinsuke Shimojo. (Search for "Other strong ecological relationships can be derived with the help of some defensible empirical assumptions concerning the relative probability of L, T, and X junctions").

So images of objects have a certain frequency distribution of junction types. It seems that written words might have a similar distribution. And indeed, Changizi and colleagues say they've shown that they do. For the frequency graphs, see Changizi's short write-up "The Topography Of Language". And that's why we're such excellent readers. Writing looks like natural scenes, which our visual systems are superbly evolved to recognise.