First we have lexis, or vocabulary which contain informational content to the listener. We define symbols such as "apple" to mean a certain type of fruit.
Secondly there is syntax, a set of rules (many implicit) that define how words are combined to form meaning.
Then there is phonetics, which dictates the sounds that constitute words (infrasgemental) and whole sentence or phrasal intonation (suprasegmental). Phonetics contains information that written language does not, such as emphasis, emotion and the speakers intentions.
In language study, most learners and teachers focus on these three. The idea is that if enough words are known, if the rules and pronunciation are practiced until they become automatic a person will have complete knowledge of the target language.
There is a fourth component in language that is often considered minor and omitted in the classroom, or at best included at the final stages of a language program.
Pragmatics is the study of the combination of context and sentences to create meaning.
Without provided context, language is provided in sentences. With included context, it is called an utterance.
In literal situations, context seems unnecessary ("the apple is green"). Many times that is fine.
But in languages that have very different social and cognitive norms, context may be essential, even in literal situations.
In Thailand, discussing money is a normal situation. In western cultures it is not. The myriad of different cultural rules like this are difficult to learn on an explicit base. But either explicitly or implicitly they form a part of the language that cannot be removed.
Here commenting on somebody's weight is fine, but only if they are of lower status or of the same level. Many social situations rely on context to determine what is and isn't appropriate to say.
Even cognitively, people can think differently in different languages.
English often uses pronouns to describe events "John broke the vase", whereas in Spanish or Japanese, speakers are more likely to say "The vase was broken", leading to a tendency for English speakers to blame others for accidental happenings.
English and French speakers think about time differently. In French, we have two main past tenses. One, the passé composé, refers to an event that happened in the past as a complete unit. Much in the way that we would use the past simple in english ("I did it"). The other, the imparfait, describes an action taking place over time. In English we can use the past progressive (I was -ing) to describe being in the middle of something when something else happens, and the imparfait can be used this way. But it also refers to something that had a long duration, such as "l’année dernière, je travaillais avec mon père", which we would translate as "last year, I worked with my father". However the French gives us an indication it wasn't a one off event, and it went on for a time, probably for the majority of the year.
Even simple situations such as the use of correct tense aren't always learned through teaching of explicit rules and the only way to learn these correctly is enough input of natural language until the learner can sense the correctness.
Once we get to ambiguous situations such as "you have a green light" as mentioned in the wikipedia entry, context becomes necessary to determine what is going on.
Most teachers and language learners probably don't argue with that. But for some reason they feel that it is an advanced subject and should be learned after vocabulary, grammar and phonetics. I believe differently.
A learner should be introduced to as much natural language as they can at early stages, even when deciphering vocabulary. The language that people are exposed to creates a memory every time. While weak, with repetition of comprehensible input, the mind uses its subconscious function of pattern recognition to create implicit rules.
It is this ruleset that gets created by the mind that the learner refers to when they seek to produce language. If high level of learning is to be achieved, the learner needs to build an accurate ruleset as early as possible. If the learner is pushed into production too early, they risk transferring the ruleset from their first language onto the language they are trying to learn.
This is how you get a person who speaks English like a Korean, or who speaks Thai like an American.
Once the ruleset is created (or transferred), it is very hard to replace, or perhaps cannot be replaced, depending on who is asked.
This is the reasoning behind ALG as a foundation method. And while some may not agree with the input only approach, serious learners of any language should at least try and maximize their input at the earliest possible stages, if they want to get to a native level.
