On 2007-06-03 08:27, Sean Whalen wrote:
> Does anyone know how he explains dehí, dhatté,
> dhi:ks.ate, hyrst, wrist, caestus, come:stus, etc.?
> Not to mention that in Iranian the same changes occur
> after x,>0, so dentals that couldn't be affricates in
> PIE even by this theory still cause the same changes
> as in original clusters.
The <dHatté> type has been discussed by Schindler. {dHad-} consistently
functions as the allomorph of {dadH-} before obstruents. On the other
hand, {dugH-} + -té gives <dugdHé>. But inconsistencies of this type may
be of relatively recent origin. We have archaising Rigvedic forms like
<aduks.at> 'has milked' or <bubutsati> 'wants to be awake' beside
innovated <adHuks.at>, <bubHutsati>, with analogical "aspiration
throwback". The archaic type survives only partly in the oldest layers
of Vedic.
In reduplicated forms (and in second members of compounds) the loss (or
non-vocalisation) of a non-initial laryngeal goes back to PIE, so
there's no reason why we shouldn't expect affricates in forms like
*dHe-dH(h1)-t- (Av. dazd-).
<dehí> must be due to dissimilation in the sequence *dHe-d[z]H-dH- >
IIr. *d(H)azdHí, with too many voiced aspirated dentals (Hoffmann's
explanation).
Piotr