Imagine that someone gives you a cake and tells you to figure out how to make the same cake. A trained chef might be able to do that just by looking at the end product, but most people wouldn't stand a chance. Let's say they then give you all the ingredients you need to make the cake and then ask you to do the same thing. Making this cake would be cake (painful pun intended) for a trained chef, but the average person still wouldn't be likely to make anything close to the original cake. Now let's say they give you instructions for making the cake as well. The chef could probably pull off the exact same cake and the average person could probably get pretty close as well.
Now replace the cake with a language, the trained chef with a polyglot, ingredients with vocabulary, and instructions with grammar rules, and it holds up just the same; you need the grammar rules to bring it all together, and getting them quickly will help you bring it all together quickly.
I noted in that the process of improving your language ability is fundamentally the same for each component of the language. The first two steps consist of identifying something you don't know and then figuring it out.
In , I argued that his method for learning grammar is slower than mine. His approach to figuring out the things he doesn't know is one of the main reasons I think his method is slower.
Let's illustrate this with some imagery.
Let's say a little circle represents an exposure to a grammar rule when using the language, and let's color code them as follows:
Let's say the first circle is your very first exposure to the grammar while reading or listening, and let's say it takes five meaningful exposures to commit the rule to memory. If , you won't have any red circles in your timeline; you'll jump right into getting the five meaningful exposures you need to commit it to memory (i.e., the yellow circles).
Steve says that he'll sometimes read the grammar at the outset, but that it doesn't stick when he does that (which makes sense, since passive exposure has a weaker effect than active exposure in terms of imprinting a memory on your brain). He also says that he doesn't really look up grammar points he doesn't get, but rather simply tries to puzzle them out from the exposure he gets. Let's say that, using this method, you need five exposures before you successfully puzzle something out (i.e., the red circles). It is only once you've gotten through this puzzling-out stage that you'll move onto the meaningful exposures you need to commit the grammar rule to memory.
I certainly do spend more time up front on grammar than Steve does, but I can't imagine that the time I spend per rule at the outset is more than the time Steve spends puzzling out the rules during exposure. How could it possibly take longer, on average, to understand something when it's written out right in front of you than to understand something when you have to figure it out from context?
Finally, like the chef in terms of the cake, Steve is a very advanced language learner. There probably aren't very many aspects of grammar that he hasn't seen before, so if it takes Steve five exposures to puzzle out a grammar rule, it might take you ten or twenty. I think Steve himself could learn grammar more quickly by focusing on it up front, but the benefit of focusing on grammar up front is even stronger for those of us who don't know eleven languages just yet.