26 Charles E. Bennett, 



^ ApLara'yopai tco 'OvaaifOiKco Berl. Phil. WocJi., 1887, No. 52, 

 1644. " Kii^hriii) ixiaadroi CoLL. 126, 2 and rw a{v)dpoi7r(o 

 126, 3 are very doubtful. 



An examination of the above instances almost tempts to 

 the conclusion that -« originated from -wi before vowels. Re- 

 jecting the last two forms as uncertain, all the others accord 

 with this inference, except too dicot Berl. Phil. Woch., 1886, 

 No. 42, col. 1323 ; TOi Maytpio) CoLL. 1 20, 3 ; 6eui too 74, i. 

 But forms in -wi also occur quite numerously before vowels ; 

 e.^. TWL ^ A\a{fji)7rpij ciTai CoLL. 60, 8 ; tcSl tXei 60, 9 ; tcSc 'HSa- 

 \if]ji 60, 31 ; tSc 'ATreiXcovt Berl. Phil. Woch., 1886, No. 42, 

 col. 1323, so that the existence of the law suggested cannot 

 be established. The absence of such a law in the occurrence 

 of -d and -di (see a, above) is also opposed to its existence here. 



4. On av by contraction from -do, see § 14, 4. 



14. 



Contraction of Vozvels. 



1. a+€ gives d in IjaaOat (for Ijdeadai) CoLL. 60, 3 and by 

 crasis in raTrt (for ra eVt) ^y, 2. 



2. d+o. Av\dovo<i Coll. 63 ; Stfiaov 69; Aaotpa 83 are all 

 too uncertain to be considered here. 



3. d+w. ^A^po9do)i Coll. 129, 130; 'Amw 97; ti/xco 69 

 are all uncertain. 



4. d+o. Final -do in the gen. sing, of masc. -d- stems (cf. 

 Homeric 'ArpetSao) contracts to -av as in Arcadian {cf. Arc. 

 ^ATToWcoviSav Coll. 1231, B, 16), vis. in 'ApiaTayopav Coll. 

 28 ; ''ApicTTijau 20, i ; %eiJiiav 66 ; ^iapaKav 29 ; NaaLcoTav 

 21, 2; 'Ovaaayopav 60, I, 22; XTacrtjav ly, I ; Ti/xayopau 

 Berl. Phil. JVoch., 1886, No. 41, viii. ; ^Ovaaayopav ibid.,x. ; 

 "^Taaayopav iv. ; IJvvTayopav Bc?'l. Phil. JVoch., 1 886, No. 51, 

 XV. ; Tifj,ay6pav ibid., xvii. 



Kvn-payopdo CoLL. 79 and l^ajaTiado 58, however, remain 

 uncontracted. Deecke on the latter regards the termination 

 -ao as also diphthongal, which is perhaps correct. Cf. Ionic 

 aoTol<i, TaoTU, auTov (for avToi<i, etc.) Cauer, Delectus^ 510. 



156 



