The key insight is this: language usage and understanding is a subconscious process. This is obvious but, I believe, has notable consequences. From this fact alone we can see why it is more accurate to call language a skill to be improved rather than a body of knowledge to be learned. Like any subconscious process, gaining increased ability is also a subconscious process. : "Learning a language" is much more similar to improving ability at a video game, or ability to draw, or ability to kick a soccer ball, than it is to, for instance, memorizing the names of body parts. By no conscious action can you directly increase your skill in a language. However, it is apparent that skill in a language can be increased. It must be increased by subconscious action. Now, it is, in theory, possible that our minds would be able to organize and assimilate explicit grammar instruction and expediently turn that into increased language ability. It is my observation that a wealth of anecdotes and even research has shown that this does not occur. Intuitively, also, it is absurd; there would be no evolutionary advantage for such an ability to have been developed. : Explicit instruction in grammar is not "inefficient"; "ineffective" is closer to the mark: it does not increase language skill whatsoever. The fundamental goal of "learning a language" is to have your speech approach that of the average person of a certain group, and the only way you could do that is by mimicking the members of that group, because a language is generally an arbitrary thing, and is not based on any particular principles that could be reasoned from. So the only plausible action of improvement would be to hear [reading is an indirect method of hearing] the language being spoken (there's no reason the brain would have faculties to interpret explicitly described grammar into its model of a language). Obviously, these utterances must be understood by the listener, because they match "what was said" to "what was meant". : comprehensible Input. Now, Input can be made comprehensible through a number of means. The main ones would be (A) explicit grammar instruction and bilingual vocabulary glosses; (B) having the meaning of the text being told to you in a language you already understand (i.e. parallel texts); or (C) having abundant context for the statement such that its meaning is inferrable. These things can be very useful, but it is good to remember they do not directly increase ability in a language, but pass through the layer of making Input to be comprehensible. As far as I can tell, there are only two determinants of the quality of input: (1) how closely it matches the aforementioned "average person" of the group you wish to emulate the speech of, and (2) how well you understand its meaning. Naturally, most everyone would naturally wish to hear material produced by a proficient speaker (usually a native), so (1) rarely becomes an issue. It's worth noting that method (C) usually produces by far the best understanding of meaning, followed by method (B) (assuming a skilled translation) and finally by method (A) (in practice, method (A) is almost always accompanied by (C), which does improve it somewhat). On the other hand, (B) is by far the easiest method and you could get through much more material in the same amount of time, compared to (A) especially, but also in many cases to (C). The balance between quality and quantity seems hard to strike, and probably varies from person to person : Grammar-translation emerges as a loser. There is no hard line between a "native language" or "second language". Both are languages which you have varying acclimation to. The only difference is that most people are weaker in the second language - meaning they have an understanding of a lower amount of vocabular and grammar structuers, and their pronunciation is further from that of the average native speaker. Plenty of people report thinking primarily in a language which was not the first. Plenty have learned a third language using a second language. Most often these cases involve English as a second language. Many people have had their native language skills deteoriate to a level below their second language skill. There is an idea that "All native speakers of a language are perfect in their usage and understanding of the language". But this cannot be the case, proven by any time a native speaker speaks a sentence another native speaker does not understand. The "native speaker" as a benchmark is simply a contrived convenience; there is a desire to define "the language" and that is the only reasonable way to go about it. However, the language itself is amorphus and cannot be precisely defined. You could say each person carries about one or more unique "languages" which they use to interact with and interpret other humans. The connotations of the words of a language are certainly part of that language, and each individual carries different connotations to different words. From this perspective, the original statement could be accurate, with the caveat that there is exactly one native speaker for any language.