English has two levels of definiteness. Definite signifies that the referent is (expected by the speaker to be an) established referent in the mind of the listener.
A third degree exists in some languages, viz. specific: an established referent in the mind of the speaker
In English, "specific" sometimes is realized identically to indefinite: "I am looking for a car" (this can be either me looking for a new car, I have not yet decided which), or I may be on the search for a very specific car, I am presenting the information so as to establish a referent in your mind. This could be followed by "The car was last seen on this very island." - the referent has been established in the listener's mind.
Other languages have no definite, only specific - anything that is definite is likely also to be definite.
There are, however, situations where definite forms are used with specific referents despite lack of frame of reference in the listener's mind: "the tea I had this morning was pretty good.", not "a tea I had this morning" or "some tea" or even just "tea". It seems restrictive attributes make specific meaning take on definite marking. The speaker does not even have to assume that the listener knows of his matutinal tea habit.
Could we draw the line elsewhere?
1) Make the marking more consistent
Naturally, we could consider the restrictive attribute by itself to be sufficiently specific that a definite article is superfluous, and have "a tea I drank this morning was pretty good" be used when it is specific. This seems even more likely if the language has a visible distinction between restrictive and descriptive attributes.
2) Have different rules in different syntactic contexts
Some languages have definiteness marked only in some contexts - c.f. the Turkish direct object rule. Here, I see several interesting possibilities:
- First person subjects license specific marking on objects and other NPs in the VP.
- Split ergativity, in which either the ergative or accusative side of the system has a case distinction that marks for specific/definite distinction. This might be triggered by person (see previous point), or various environments such as subclauses (maybe specifically narrative such).
- Maybe specific and definite nouns interact with congruence in different ways. This may restrict the environments in which it is explicitly marked:
- Maybe only subjects (or only objects, or both but not other constituents) have verb congruence that permit for this distinction?
- Maybe the marking on the adjective also is two-fold, but splits the difference differently. Thus red.def house.def is definite, but red.indef house.def is specific. (Here, "red.def house.indef" would seem like an attractive solution as well, if we assume Adj N word order.)
3) Multiple levels of specific-definite-contraspecific and ways in which the specificness and definitenesses of different speech participants interact in marking.
It is conceivable, that a speaker might want to communicate that he does not have a clear idea yet of the thing the listener has spoken of, and so could mark the lack of understanding as [+definite -specific]. In such a language, clearly, both specificity and definiteness need to be marked independently - but potentially, it could be marked independently in a way that isn't always visible or always clearly distinct. Consider, for instance, a system where adjectives mark for indefinite, specific-or-definite, whereas nouns only mark "specific-or-indefinite" vs. "definite". The normal "specific" combination would thus be "specific-or-definite" adjective but "specific-or-indefinite" noun. However, in that case, an indefinite adjective combined with a definite noun would perchance convey this confusion. The locuses needn't be nouns and adjectives, could as well be verbs and nouns or other carriers of congruence. Any ways, the adjectival congruence solution is nice because adjectives are often optional.
4) Have different rules for nouns of different topical salience
One could imagine a system whereby nouns that are topics have more levels of distinction.
5) Have different rules for nouns of different noun classes or number
Plurals or inanimates or mass nouns might very well differ. The difference may be a thing that has purely cultural origins, or may be the result of sound changes eliminating the distinction for some forms.
6) In some languages with articles, there are contexts where no article is used. One could consider having articles dropped whenever there is tension or uncertainty regarding [?specific ?definite], or whenever unusual combinations such as [-specific +definite] appear.