Hallo!

I have just found out something about the numerals of Old Albic.

The numbers 1-12 (Old Albic numerals are base 12) are:

_sama _1; _tovu_ 2; _theri_ 3; _letvi_ 4; _phendi_ 5; _tøthri_ 6; 
_hesphi_ 7; _tølti_ 8; _terthri_ 9; _tøphni_ 10; _sendi_ 11; _rasta_ 12.

These are inflected like animate nouns and agree with the counted
noun in case.  The final vowels are of course number markers
(_-a_: singular, _-u_: dual, _-i_: plural; _rasta_ `12' is a
special case, it might be better translated as `a dozen').
You may have noticed that the numerals from 3 to 11 all have
front vowels in their root syllables.  This is of course to be
expected as the /i/ of the plural suffix umlauts the preceding
vowel.  Indeed, for some of them, forms with back vowels are
well attested in dialects or derivations, and the phoneme /ø/ 
is a product of i-umlaut anyway.  So, we have the following roots:

*sam- 1; *tov- 2; *thar- 3; *latv- 4; *phand- 5; *tothr- 6;
*hesph- 7; *tolt- 8; *tharthr- 9; *tophn- 10; *sand- 11; *rast- 12.

Of these, *sam-, *tov- and *thar- are unanalysable, but resemble
their Indo-European counterparts *sem-, *duw- and, less perfectly,
*trei-.  It is tempting to interpret the element *-tv- in *latv-
as a kind of zero grade of *tov-; but what then is *la-?
A possible connection is to _lothu(m)_ `pair' which is of course
*lath-u(m).  So *latv- could be from *lath-tv- `a pair of 2s'?

The numeral root *phand- `5' distantly resembles PIE *penkWe-,
but only the first three segments match, and *phand- is generally
considered related to _phañ_ `hand', apperently via *phañ-d- >
*phand- with assimilation of the velar nasal.  The root *hesph-
`7', in contrast, has so far defied analysis.

`6', `8' and `10' form an interesting group.  They all begin with
*to-, which, as all these numerals are even, could be related to
*tov- `2'.  If we split off this initial *to-, we get the remaining
forms *thr-, *lt-, *phn- which could be interpreted as zero grades
of *thar-, *latv- (with loss of the semivowel at the end) and
*phand- (with loss of *d), hence, these are `2*3', `2*4' and `2*5'.
Similarly *tharthr- `9' <  *thar-thr- `3*3'.

Remain *sand- `11' and *rast- 12.  The latter is not analysable,
but *sand- might be from *sam-da- `one off [twelve]'.

Finally, there is the word _stala_ `0'.  This is a new formation
connected with the invention of the zero in Classical times,
derived from the word _stal_ `empty'.

Greetings,

Jörg.