Why do people make grammar errors? In a general sense, errors are of two types: errors of competence and errors of performance. Errors of competence are the easy ones to explain; they’re mistakes made because one doesn’t know the correct form. I’m well familiar with these mistakes from speaking in my Spanish classes last year despite not knowing what prepositions go with what verbs and so on.
Errors of performance are a bit more interesting. Performance errors are those where one does know the correct form but still makes a mistake. For instance, I generally know the rules of English grammar. Yet when a helpful reader went through the blog archives and copy-edited for me, the list of typos and other errors he’d found were staggering. For most of the errors he found, I immediately agreed that I’d made a mistake. So why would I have screwed these things up if I knew they were errors?
Well, language happens in real time, with real-time constraints and pressures, such as having to convert from language in the mind to mouth/hand movements. That makes the task difficult, and difficult tasks are open to performance errors. Language production is difficult enough a task that immediate perfection is usually a pipe dream. The number of performance errors can be reduced if one has a chance to go back and edit, but even then some errors might slip through.
In general, spending more time on production (let’s include editing time in production time) decreases the frequency of errors. Faster production will have more performance errors, slower production will have less. This is an example of what’s known as a speed-accuracy tradeoff, and here’s a schematic of it from an MIT course:
Since language is so automatic, it might be difficult to think of changing the speed of production to affect accuracy. So to make it a little clearer, I’m willing to return to a bane of my elementary school career: Simon Says. The game is pretty simple, where you do what you’re told, but only when it’s prefaced with “Simon says X”. When commands are given out slowly, there’s almost no danger of screwing up and following a bare instruction. Low speed, high accuracy. But as it gets faster, you’re doing things sometimes before you fully process them, and that means that sometimes you’ll jump at “Jump!” before realizing that there was no “Simon says” before it. High speed, low accuracy.
Speaking or writing is much the same. If you’re going slow, you’re not going to make that many errors (unless you’re going slowly because your focus is divided or the task is difficult). If you’re going fast, you’re more likely to screw things up, even when you know the rules.
So if slowing down reduces errors, why don’t we all just slow down? As my friend Dan put it, you can speak really slowly and make (almost) no errors, but your audience will walk away before you finish your sentence. Instead, the intelligent speaker/writer aims for some reasonable combination of speed and accuracy. What’s reasonable changes due to external pressures, such as the urgency or importance of a situation. A shouted “Move!” is appropriate if a car is barrelling down on a friend; “You should take a few steps back to get out of the way,” may be more appropriate when the friend is in the path of a crawling steamroller. A quick text to a friend can have errors if it saves time (and remains comprehensible); an important email to a client is worth the extra time to reduce the errors. Problems arise when someone thinks they have more time than they actually do or they need less accuracy than they actually do.
The point of all of this is that a rational speaker or writer won’t necessarily aim for maximally grammatical production, unless the task requires it. Instead, the rational producer aims for maximizing grammaticality without spending excessive time. If there is a better way to spend one’s time and the production is good enough, a rational producer should send it off as-is. As a result, we shouldn’t think that someone making a grammar error means they don’t know the grammar, or that they don’t care. If speed is the primary concern, one has to accept some errors. If errors can’t be tolerated, one has to accept slower production.