In his paper at:
www.unc.edu/~melchert/anzan.pdf
Melchert attempts to explain Hittite words ending in -anzan- with nom.
-anza(s) as derived from *-tyo-stems.
Much of it is convincing, but there are some problems. First, just
because sumanza(n)- doesn't come from *syu:m()n- (any more than any of
the others are from -m()n-) doesn't mean that it doesn't come from a
derivative of *syu:- or have to do with binding because it meant
bulrush. Many plants have multiple uses and characteristics and yet
are still named only after one of them.
Second, his explanation of the alternate forms of the nom. -anza(s)
is, as he admitted, uncertain. The only way to account for the forms
is to start with immediately previous nom. -anza and stem -anzan- with
the -s a recent analogical "correction" to a more common ending.
The explanation of their origin further in the past should take into
account the other H word with nom. -za but stem -n- : *gWanza *gWan-
'woman' from earlier *gWana *gWan- (like Luwian wana- (with alt.
wanatti- < *widatti- 'widow')).
In other words, these are from feminines in -a: with the oblique
changed by analogy with *gWana *gWan- when they still both had
feminine gender (later *gWana itself was changed by analogy in the
other direction in H (but not Luw)).
Though the fem. isn't thought to have existed in Anatolian, there
are direct parallels:
*sentontyax > L sententia 'way of thinking, will'
*s()tontyax > H istanza(n)- 'mind, soul, will' by haplology of en-en
*xn,xtnix 'female duck' (with -nix < *potnix 'lady', most have x-x
and/or n-n dissim.)
*(x)n,xt(n)ix > *naxtyax > G ne:^ssa 'duck'
*(x)n,xtnix > *naxtnyax > *naxantya: > H lahhanza(n)-
(probably the H became the only fem. in -i and so added normal -a to
make -ya like Latin -i(s) / -ia, etc.)
Others are derived from verbs essentially as L sententia:
*syuxW+ 'sew, bind' > *syuwont- 'binding' > *syuwontya:- 'binding
thing, cord' > sumanza(n)- 'bulrush'
*kal+ 'rise, grow' > *kalont- 'growing' > *kalontya:- 'growing thing,
plant' > kallanza(n)- 'kind of plant'
As sketched earlier, when the number of feminines decreased, perhaps
mainly to the productive -entia ending and 'woman', their paradigm
would have been altered by sound changes so final -a: > -a. The
relation between *kalont- 'growing' and *kalontya:- 'growing thing',
for example, would have become:
*kalants ... *kalantsa
*kalantan .. *kalantsan
*kalantas .. *kalantsas
Whether to make the derivatives less similar to the derived or just
by analogy between fem. forms, -n- was added from *gWana *gWan- to the
obl. of all remaining fem.
*kalants ... *kalantsa
*kalantan .. *kalantsanan
*kalantas .. *kalantsanas
Now every case had an extra syllable in the derivatives (although
whether that was the intent of the analogy, I can't say).
Later, when fem. in -anza were the only other fem. left, *gWana
(Luwian wana-) itself was changed by analogy in the other direction to
H *gWanza.
Still later, the fem. began disappearing as a category, and -s could
be optionally added to the former fem. in -anza < *-ontya:, but not
*gWanza itself, as 'woman' was likely the last word 'defemininized'.
Previously, I didn't think any traces of the fem. existed in
Anatolian; my old derivation below no longer is needed.
--- In cybalist@yahoogroups.com, "stlatos" <stlatos@...> wrote:
>
> *sent-tu- > L sensus, H istanza- 'soul'
>
> *sent-tu-
> *sten-tu-
> *sten-tw-V
> *sten-tsw-
> *sten-tsy-
á é í ó ú à è ò ì ù È Ó
æ ø å ä ë ï ö ü ő ű þ ð ç