[DiogenesKynikos]

The range of words you learn with Pimsleur is limited because the method makes sure you actually retain every word it tries to teach you. Having a solid grasp of a small vocabulary (a few hundred words) is an excellent way to start off with a language.

> But I think you won't be able to really grok "random" speech before having a good grasp at pretty much all the vocabulary you would encounter.

But you will be able to grok small conversations about subjects you're familiar with. You can expand from there. When you try to learn a new language, you should be speaking from day one, no matter how limited the range of subjects you can speak about.

[bayesian_horse]

Yes, which is why I say the phases overlap. I still think theory should dominate the first phase of learning a language, followed by mostly reading.

[DiogenesKynikos]

Having learned a few languages with different methodologies, I strongly disagree. The most effective introductory method I've encountered so far is Pimsleur, because of its focus on listening and speaking at the early stages.

I've seen too many students who learn using the traditional theory-heavy approach and who, after years of study, are like deer caught in the headlights when they're confronted with the actual spoken language. They're also often saddled with poor pronunciation forever (which really degrades native speakers' perception of your language proficiency, regardless of how fluent you are).

Some theory is helpful, particularly for languages with lots of grammar (i.e., Indo-European languages with case, gender, agreement, etc.), but like playing an instrument, learning a language is 90% practice.