Quoting Steg Belsky <[log in to unmask]>:

> On Mar 12, 2005, at 9:29 PM, Chris Bates wrote:
> >> There is a universal phenomenon, where sometimes definite articles
> >> lose
> >> their power to definitize.  Because of this there are those who claim
> >> that what is today an indefinite article (Arabic |tanwiin| -n,
> >> Akkadian
> >> -m) once was a definite article.
> > I've never seen an example of a definite article turning into an
> > indefinite article (I don't speak Arabic though). It's true that
> > definite articles often lose their definiteness, but this isn't the
> > same
> > as turning into an indefinite article. The tendency in many languages
> > is
> > for usage to spread until the article indicates specificality rather
> > than definiteness and from there turn into a general NP marker. At this
> > point it covers both indefinite and definite meaning, but it hasn't
> > turned into an indefinite article.
> But then what if a new definite article develops, from a deictic or a
> pronoun or whatever?  Then the old definite article might become
> understood as an indefinite article.

This happened to one of my conlangs, namely Telendlest; the old definite article
was generalized to most NPs (the chief exceptions being proper names,
"adjectival" genitives, and NPs with a demonstrative), and then a demonstrative
got demoted to definite article, with new demonstratives being derived from the
same root. Since the old def article now only occured on indef NPs,
reinterpretation as an indefiniteness marker was more-or-less automatic.

This causes no end of confusion to speakers of related languages that retain the
old article as a definite one.