² The author has in preparation a full description of Romanian, based on the same material.

³ By standard Romanian is here meant the dialect of Bucharest as used by the two informants whose speech the author has analyzed: Iris Barbura (B), a native of Arad, Transylvania, who has spent most of her adult years in Bucharest; and Iulian Prundeanu (P), a native of Bucharest. Observation of the speech of numerous other Romanians of various ages, natives of several provinces, has revealed no important structural deviations from the dialect of B and P.

⁴ Since all the forms in this paper are merely cited words, not communicative utterances, the pitch and juncture phonemes need not be included in their notation. The citation contour being uniformly /(1) ... 2' ... 1./, /tátə/ stands for /²tátə¹./ 'father', /kibrít/ for /¹ki²brít¹./ ‘match’, and so on.

⁵ There is a residual class of uninflected words.

⁶ The parentheses signify that the definite inflection occurs without the oblique, but not conversely.

⁷ These phonological words do not inflect to produce the total paradigm which defines a noun. /deobičéj/ is matched only by /deobičéjurj/ ‘of customs’; /umpriéten/ does not inflect at all. On the other hand /običéj/ ‘custom’ and /priéten/ ‘friend’ are fully inflected.

⁸ All the examples used in the rest of this paper are included here, alphabetically arranged within each group. The accompanying glosses will not be repeated in subsequent citations.

⁹ Informant B says /fíkə/, as in her own and P's /pisíkə/ ‘cat’; but she has /íj / elsewhere in ber speech.

¹⁰ Stress is nonfinal in all these bases, and therefore needs to be written only in those which contain more than one nonfinal nucleus (i.e. vowel, or cluster /ea oa/).

¹¹ It would be possible for the -u- to alternate to final zero in types like fiu- or riu-; but the pattern is consistently /u ~ w/ after any stressed vowel except in skiu- alone, which has the bare base /skí/.

¹² Save in the single instance of /ów ówə/, where, instead of dropping, the base vowel alternates to /w/ and thereby brings about, because of the limitation on /we/, a unique manifestation of the plural -e- as /ə/.

¹³ On the other hand it would not be automatic before a morphophoneme -j-, for /Vj/ occurs.

¹⁴ This is merely a free alternant of oáspete-, with bare base /wáspete/. The behavior of the morphophonemic cluster -oa- is treated below.

¹⁵ The alternation /ə ~ e/ of -ə- takes place also in unstressed positions, posttonically in polysyllabic bases, as in numəruə- /númər numere/, tínəru- /tər tínerj/. (The morpho-phoneme -i- is explained below.) After a velar consonant /ə/ remains unchanged, as in flə /flákə flərj/, strígə- /strígət strígəte/.

¹⁶ Cf. also oasp[et]e-, oam[en]u-, cited in §4.

¹⁷ /váčij/ means, in isolation, ‘to the cow’.

¹⁸ Regarding the /w/ of /kafeáwa/, see fn. 19.

¹⁹ The /w/ which occurs in the definite forms /kafeáwa/, /steáwa/, etc. can be classed as a unique manifestation of the base consonant -L-, appearing instead of the expected zero before /a/ and providing a contrast between the bare base /kafeá/ and the definite /kafeá+a/. Because of the limitation on geminate vowels, there can be no */kafeáa/. The unique ziLə ^v- could, phonologically, have the definite form */zía/, but it patterns with the other members of its base type (subtype) in having /zíwa/.

²⁰ One ə-base, namely tatə-, has either the standard shard əa- or the special shard əu-, appearing as in (4): tatə.L.a.(.j) = /táta tácij/ or tatə.L.u(.j) = /tátəl tátəluj/.

²¹ If shards were present, we would not expect them to have zero alternants before oblique -or-, for both /eo/ and /io/ occur.

²² Also in the unique case of taə- (or taəu-, cf. fn. 20) : /tácij/.