[troad]

You are entirely correct in your criticism of Rosetta Stone, but LLPSI genuinely takes a very different approach. (I've done both, though for different languages.) LLPSI is carefully structured, precisely to teach things like declension and conjugation. As an example, this is very first line:

Rōma in Italiā est. Italia in Europā est. Graecia in Eurōpā est. Italia et Graecia in Eurōpā sunt.

This is not what most people mean by immersion. It's not a natural dialogue you're expected to understand by osmosis, it's an extremely carefully designed series of sentences aimed at manually bootstrapping your Latin. The book contains grammar explanations - wholly in Latin - from chapter 2 onwards. Every chapter has a grammatical concept it's designed to illustrate, and it manages to introduce fairly complex ideas - passive conjugations! ablative declensions! - in a deeply intuitive way, entirely in Latin. It's an incredibly satisfying course. If you have any interest in Latin at all, or are just looking for a New Year's Resolution, I would heartily recommend it.

[RheingoldRiver]

That's interesting actually. It suggests that a major reason this works is because of its cognates with English that give you a starting point. So for example, a native Korean speaker would really struggle to learn from this curriculum, and the strategy wouldn't work very well to learn Korean. I wonder if it's possible at all to convey enough with pictures to design a curriculum like this when you don't have the help of cognates.

[troad]

Cognates and illustrations definitely assist, and the book uses both freely. Chapter 1 begins with a map, for example. It’s a very clever strategy, one especially well suited for teaching Latin to people already speaking an Indo-European language.

I imagine it would be less useful to people who speak a language that doesn’t share the features of Latin. A map isn’t going to be of much use if one’s native language doesn’t have the concept of plurals, and one therefore struggles to comprehend the est / sunt distinction altogether. Then again, English doesn’t decline nouns or conjugate verbs (for the most part), and English speakers tend to be big fans of LLPSI. To its credit, it goes slow, and reinforces its lessons very well. The rest of chapter 1 is basically spent on variations of the above grammar point, introducing new nouns but reusing those two verb forms again and again in different variations, illustrating their use and how they contrast.

To your question, I also wonder how well this kind of bootstrapping approach could be achieved without relying on language similarity. The trick seems to be to tap into pre-existing adult skills or knowledge - e.g. logic, map reading. I think this is what sets LLPSI apart from “immersion” approaches like Rosetta Stone, which are based on the erroneous assumption that adults (do / should) learn language as toddlers do.

[gnubison]

The book also includes a map of the places mentioned, at the time of Classical Latin that the narrative is set in. I’m only about half way through, but there really aren’t that many cognates. An aha! moment every so often, sure — like silēre -> silēns -> silentem -> silent — but it’s not like I read through and it only makes sense because of cognates.