The Phonetic Law -CnT# > Cr(T)#
So I have established the phonetic law "-CnT# > -Cr(T)#". "C" is the symbol for any consonant and "T" means the phonemes "d, t, dh" equally. So -n- and (later) -r- in this position are sonantic. "#" symbolizes the end of the word. In this position "T" shows a tendency to disappear which is symbolized by the brackets.
The reason for this law is less effort in speech as can easily be tested. Sonantic -n- seems to have been less stable than sonantic -r- and often developed into a vowel in Indoeuropean languages. The phonemes a, r, l, n, d, t and dh are related because all of them are produced by the same mouth position i.e. the tongue in flat (not arched) position. The development from -n- to -r- means first of all loss of nasalization and therefore the attainment of a higher degree of phonetic relationship to the following "T" which is not nasalized. The process may be interpreted as partial assimilation.
What are the consequences of this law? First of all the
ending -r of the heteroclita may be analysed as -n + -d which developed
to -r + -d and finally to -r. That means that there is no stem change at
all: heteroclita are neuters of the n-declension with the same gender marker
as pronouns. There is no difference between the inflections of nouns and
pronouns! The heteroclita disappear in modern Indoeuropean languages but
Hittite and other old Indoeuropean languages show that this class was formerly
wide-spread. On the other hand neuters of the o-declension are an increasing
class. I assume that this class is influenced by the animate gender where
the marker -m generally occurs. The remaining class of endingless neuters
might have been influenced by the heteroclita which seemed to be endingless
as soon as the previously described phonetic law had taken effect. Another
position where this phonetic law must have taken effect exists: The third
person plural of the "secondary" verbal endings of the active voice.
This ending is -nt and must have developed to -r(t) immediately after a
consonant. I think that the third person plural of the perfect ending
-r is of the same origin. The secondary ending -r does not occur at
"thematic" stems and at "athematic" stems which show the ending -ent which
seems to be the alternative to avoid the ending -r in certain verbal classes
(unless due to accent). The ending -r occurs at verbal stems of the "third"
Indoaryan verbal class which show present reduplication and no thematic
vowel e.g. 3. plural imperfect abibharur < ébibhernt vs.
3. plural present bibhrati < bibhrnti vs. 3. singular imperfect abibhar
< ébibhert vs. 3. singular present bibharti < bibherti, cf.
the verbum of the "first" (thematic) class abharan < ébheront
vs. bharanti < bheronti vs. abharat < ébheret vs. bharati
< bhereti. It is clear that a strong tendency to reestablish the old
ending -n(t) existed, due to the thematic verba in -on(t) and the primary
ending -nti which took place in Greek and also in Gatha Avestan. A similiar
tendency seems to have been responsible for the lack of the ending -r
in neuter nouns of the nt-declension. But perhaps the Hittite verbal
substantive in -uwar which corresponds with the infinitive in -uwanzi shows
the law (-uwar < ..nt = nominative against -uwanzi < ..nt-i = locative