I was reading a little style guide on Slate Star Codex. Now truth be told, I generally find this kind of “don’t say this, say that instead” style guide somewhat patronizing and quite irritating (more of a testament to my own rebellious spirit than any indictment of any author) and unhelpfully unnuanced (a more practical complaint), and my first instinct was to want to argue this lack of nuance. On the other hand, Scott is a very skilled communicator and an examplar in how being an enormous dork need not be a barrier to popularity, and there is a more helpful general principle hidden in these rules.

The principle here is that while reading (or listening, viewing, …), people are constantly predicting what will come next. If what they read is what they expect - all good, the reading flows smoothly, and people interpret the text as saying what they already thought it was saying, which is low effort and comfortable. If, on the other hand, they encounter something unexpected, this will stand out, draw their intention, be interpreted as meaningful.

Staying on the level of single words and turns of phrase, in practice what this means is that when you use an unusual word or phrasing instead of a more conventional (to the reader! “normal speech” is audience-relative!) synonym, it will be taken as deliberate and specific; the reader will interpret your choice to use that word as you having searched for the right word to use because the specific meaning matters.

Often, this is not what you want. In scientific discourse, precision is highly valued, and so scientific writing has a house style of using carefully chosen, specific words. In normal everyday prose, however, this amounts to information overload. Even if the length of the text ends up the same, by choosing unexpected “pokey” words, you are preventing the reader from rounding your message off to their own everyday working set of concepts. In effect, you’re making your message less compressible for them. Try and do this sparingly! All of this is simply a special case of the commonsense principles of getting to the point and avoiding extraneous detail.

I am fairly sure this is all supposed to be strongly related to the linguistic principle of markedness, but the wikipedia page on markedness is too technical for me to care enough about parsing and I get the impression that “markedness” is a broad multi-dimensional idea of which this post is a specific instance and my perspective on the topic is too superficial to point at the specific thing I am thinking of.

Having arrived at a nice concise principle of communication, let’s take a step back and generalize a bit, because I think this idea of the brain as constantly predicting sensory input and responding to surprises is useful and interesting. Specifically, while writing this it called up something I have read about schizophrenia. In a nutshell, schizophrenics commonly experience something what is called “delusions of reference”, in which they interpret innocuous things (e.g. newspaper headlines, things said on radio) as having special meaning to them. In some theories of brain function, there is an explanation for this that goes as follows: the brain is constantly predicting upcoming stimuli. In people with schizophrenia, this sometimes goes awry in a way that makes the brain flag something innocuous as deeply surprising. To the schizophrenic person, this feels as though the stimulus in question is somehow deeply meaningful to them personally, presumably in the same way that choosing an unusual “pokey” word instead of a more common synonym feels deliberate and meaningful.

This is all related to the neuroscientific paradigm of predictive coding. I am not articulate enough in this to write about it at length and since all I know about this is from SSC articles I have read, I will just link those: I got the bit about schizophrenia from It’s Bayes All The Way Up and for more in-depth state-of-the-art neuroscience stuff consider Scott’s excellent review of Surfing Uncertainty.