1) Necessary features
Some things are true of all languages by necessity, where language is a discrete combinatorial system of signs used to encode and communicate complicated information about one’s world.
One person talking to another person who cannot reply in a meaningful way does not a language make. If only one person could understand a language, it would be more akin to highly systemized, individual babbling.
1.2) Total feedback
Communicating without total feedback is a notion perhaps best explained by example: a babbling infant might accidentally spew the phrase, “I am self-aware,” and yet that infant had no awareness that the phrase he uttered had meaning, much less what that meaning could be. Whereas language without interchangeability is like talking to a brick wall, language without total feedback is like a brick wall talking – keeping in mind that it has no way of assigning meaning to its utterances.
Language without specialization would be very similar to language without total feedback: if a phrase is uttered without any intentional meaning, it would not only not represent any thought of the speaker, but could possibly even misrepresent his intent.
Language without semanticity is yet another form of babbling. If one wanted to refer to a dog pouncing on a squirrel, but none of the words in the message had one set meaning, the recipient of the message could well interpret “A dog pounced on a squirrel” as “Go eat three bagels and cheese,” or whatever else they might guess it to signify.
Arbitrariness is actually a feature present in all systems of communication. Without arbitrariness, spoken language would be impossible, imagine if one were to use his larynx to produce a picture of a squid to refer to a squid. Assuming that to be impossible – because it is – this must even apply to written language. Imagine drawing a picture of total feedback, semanticity, or arbitrariness.
If the units of a language were not discrete, they would be useless, as we would be unable to separate them from each other. Practically, this would be very similar to a lack of semanticity: one may know what he wants to say, but would have no ability to do so in a way anyone else could understand.
If one could not combine elements differently, he simply would not be using language. This would be more equivalent to a chimpanzee’s innate communication. Similarly, shrieking and banging one’s head against the wall to communicate that one is unhappy does not constitute language.
2) Sufficient features
Some things are just incredibly useful features of language, and therefore evolved to be ubiquitous in natural language.
Transitoriness is simply the difference between [trænzətoʊɻij] and [tːrːːæːːːnːːzːːːəːːːːːtːoʊːːːːːːːːːɻːːːːiːːːːːj] when there is no phonemic distinction between ten different lengths.
A language without displacement would be very limited in scope and usability, but that would not change the fact that it is a language. There are likely some constructed languages, for computer or other use, that do not have displacement.
2.3) Traditional transmission
If one’s father spoke a thieves’ cant, and did not pass that cant on to his son, it would not be any less of a language. The same is also true of natural languages. It would be rather stupid to not pass natural language on to the next generation, because that would prevent communication with them, and would set human evolution back hundreds of thousands of years, but it is conceivable that a language would not be passed on by tradition.
2.4) Duality of patterning
There is no reason that the base units – phonological segments, or individual letters – could not have meaning themselves. While it would be rather inconvenient in natural language – well, perhaps not in !Xóõ which has as many as 200 phonemes – it is not unlikely to be found in constructed languages, particularly experimental ones.
A language without reflectiveness is limited in a manner similar to a language without displacement, except that it is a much smaller limitation – at least, for non-linguists. In fact, I have written entire essays, stories, and poems, and carried on many conversations, without once referring to language itself. As an example, most if not all computer languages are incapable of this.
3) Spoken language features
Some things are only relevant to spoken language, and may only be true because of human biology.
3.1) Vocal-auditory channel
Written language is emitted by a writing utensil and perceived by the visual system. Sign language is emitted by the body and perceived by the visual system. Then there are Morse code, smoke signals, and (conceivably) many other messaging systems that do not rely on the vocal-auditory communication channel.
3.2) Broadcast transmission and directional reception
This is a property of sound, which is only applicable to communication by auditory channels. It is utterly irrelevant to signed, written, or other languages.
4) False features
And some things either are not entirely true, or are true but result in the degradation of language.
Not everyone who is capable of language is capable of learning more languages. In the case of humans, as one gets older, it gets significantly harder to learn a new language, to the point (at a relatively young age, all considered) that one can no longer completely learn a new language. This may also apply to people with certain mental deficiencies – learning disorders, to be specific: while they can acquire language, acquiring another would be difficult at best.
Prevarication is actually the antithesis of language. Every instance of lying or deception degrades language, eventually to the point where it becomes meaningless; specifically, prevarication is the opposite of semanticity. Though a full explanation of this principle would be far too long for this space, the reader is encouraged to peruse the work of Immanuel Kant.