140 THE FORMATION OF NOUNS to this type and occurs most frequently in Sanskrit, but there is also a type with retracted accent associated with a curious syntactic distinction : ddtA vdsiindm but ditd vasum. It has been noticed above that specialist words with this suffix (hotar-, etc.) have also as a rule accent on the root. In Greek also there are two sets of forms, with suffixal accent, S orrjp ' giver parrjp 'goer 1 , derrjp 'established, and with radical accent, Sdyrcop ‘ giver etc. The first of these preserves the most ancient form, with reduction of the root consequent on the accentuation of the suffix. In Sanskrit the suffixal accent is preserved to a large extent, but apart from very few exceptions, e.g. drmhitdr- 4 one who makes firm ', guna is universal in the agent nouns. Its maintenance or reintroduction in spite of the basic law of apophony may be ascribed to the influence of the related neuters. This suffix is prominent in the formation of nouns of family relationship : pitdr- ‘ father ' (cf. Lat. pater , etc.), duhitdr- 4 daughter ' (cf. Gk. Bvydrrjp , etc,), mdtdr- ( mother ' (Gk. ^rrjp. Dor. fiaTTjp, OHG muoter, etc.) ; bhratar - ' brother ' (Gk. <fi par cop, pparrjp, <f>p&Trfp 1 member of a phratry Goth, hr d par, OHG bruoder , etc.), jimdtar - ' son-in-law ' (Av. zamdtar-, Alb. Sender)] yatar- ' wife of husband's brother ' (Gk. elvaripzs, Lat. ianitnces plur., O. SI. jgtry, Lith. jente) ; ndptar - 1 grandson ' (secondary substitute for ndpdt — Lat. nepos). Of these it is probable that the word for ' father ’ is an old agent noun (p-i-tdr- 4 protector ' from pd-(y)~ 4 to protect '), but in the majority of cases the etymology is too obscure for it to be pos- sible to say much with certainty. Its gradation is of the old type (cf. Gk, Sor^p, etc.) as is to be expected in such a word. Only duhitdr - agrees with pitdr- in accent and apophony ; the rest have both accent and guna of root with the exception of mdtdr-, and even here Greek has radical accent, which may easily be original in spite of the agreement between Sanskrit and Germanic. It is not unlikely that these contain some old neuters etc. : Lat. materies would be an extension of such a neuter) which were adapted when the gender-system developed. There is also the possibility of the analogical exten- sion of the suffix. This has certainly happened in Skt. ndptar- and probably in the unusually formed jamdiar- (cf. Gk. yajxfipo? son-in-law ' differently formed. The defective noun star - (instr. pi. stfbhis) , tar- (nom. pi.
