[ehartsuyker]

This is everything I hate about language learning through pure immersion (Duolingo, Rosetta Stone, most language schools). For example, I struggled for an absurdly long time with genders in German when I started because I assumed it was like French (male, female, plural). Turns out there's 4 genders, and often the female article is the same as a the plural. Once English sentence could have solved that.

[mikeash]

I think that learning the rules is an absolute necessity, even if your goal is to develop your intuition well enough that you don't need them. (Edit: I think this implies a little more than I meant to say. Being exposed to the rules is a necessity, but you don't necessarily have to go out and memorize them from beginning to end or anything.) They're important scaffolding. Even native speakers of a language need to be taught the formal rules if they're to speak well.

I'm going to make a lame attempt to coin a term and call this the Arch Fallacy, which is when you look at your goal and assume anything not present in it is unnecessary to achieve it. A completed arch contains no supports within it, but you'll have a tough time building one without any.

[frenchy]

> Even native speakers of a language need to be taught the formal rules if they're to speak well.

Native speakers are taught some rules, but many are rarely taught unless a student has learning difficulties. The adjective ordering that OP mentions is a solid example. Others include aspiration of stops; voicing assimilation in words like "bets" and "beds"; the difference between "putting up" and "putting out"; why it's (sometimes) okay to say "Hey fucker" to your best friend, but not to your teacher; and the deletion of nominative arguments in subjunctive phrases.

Second language learners often (maybe always) find more rules useful, for several reasons. However, those who learn the language in highly rule-focused manner often have difficulty actually communicating in the same way that Computer Science professors are sometimes quite knowledgeable about CS theory, but unable to produce good code. Rules are good at telling you what not to do, but bad at helping you be creative.

[mikeash]

Yeah, native speakers won't typically be exposed to the same rules, and there aren't a lot of rules taught, but they are there to an extent.

I wouldn't say that one should learn in a rule-focused manner, but neither should one learn completely by example. A mix is best, even when your goal is to have the rules fall away in the end.

[Stratoscope]

The new eastern span of the Bay Bridge is a similar example. They had to build a complete temporary bridge of iron girders to support the new roadway before the tower and cables were in place. (The bridge design didn't allow the tower and cables to be built first and then the roadway suspended like a conventional suspension bridge.)

This page has a good video of the combined structure and the process of shifting the roadway load from the temporary bridge to the cables:

http://baybridgeinfo.org/projects/sas

The background image on the page is also nice, showing all three bridges together:

http://baybridgeinfo.org/sites/default/files/images/backgrou...

[Stratoscope]

"contains no supports" (was "contains to supports")

Funny thing is on first reading, I didn't notice that at all and read it the way you meant. I only saw the typo on a second reading. Interesting autocorrect the brain does.

In any case I like your Arch Fallacy! I think I will use it.

[TeMPOraL]

> In any case I like your Arch Fallacy! I think I will use it.

I'll use it too, it's great and I don't recall any term for that particular error in thinking. I've already started spreading it (the name, not the error) among friends, hoping it will catch on.

[mikeash]

Thanks for finding the typo. Glad you like the idea.

[stevoski]

There are three genders in German. And four cases. Multiply them together, throw in plurals, add in uses of the indefinite article and the definitive article, and, well, it is far too much for a single sentence to explain.

But in principle I agree with you. We need a bit of book learning to go with the "just start speaking" approach.

[SilasX]

Hence the 3-dimensional chart for determining adjective ending -- it depends on gender, case, and the word preceding it (i.e., definite article, indefinite, or nothing). Although I found a chart on this page that "reduces" it to a four-branch flow chart:

http://www.nthuleen.com/teach/grammar/adjektivendungenexpl.h...

[tmd]

After having lived 4+ years in Germany and struggling to learn German, I would say that the only way you can learn truly by pure immersion is from your parents. In any other case you need at least some theoretical basis because no one else is willing to spend hours trying to communicate with someone who's basically babbling.

I'm using Duolingo now and after I'm finished with it I want to get some formal classes to (hopefully) get to a B1 level. I wouldn't call Duolingo immersion learning though -- they explicitly provide grammatical rules (unless you're using the mobile app which lacks some features) and it's very difficult and frustrating to try to do it with your intuition alone.

I also liked the explanation of cases and adjective endings from this site: http://yourdailygerman.com/learn-german-online-course/

[jdeisenberg]

In the language tutorials I wrote, I decided to write a whole section for people who didn't want to learn grammar. (http://langintro.com/kintro/grammar/hategram.htm) My example was Russian adjectives, where a 3 by 5 chart reduces to a 5-entry table plus two spelling rules.