¹ This study arose from a conversation with Morris Swadesh at the meeting of the Linguistic Society in 1949, and from further discussions with Samuel E. Martin and William C. Sturtevant. The description is based on my own speech and that of Lic. Carlos José Gutiérrez of San José, Costa Rica; it represents the urban dialect of the Meseta Central. which includes not only San José but also the major cities of Cartago, Alajuela, and Heredia.

I wish to express my thanks to Lic. Gutiérrez (now at the University of Pennsylvania) for the time that he so willingly gave me; to Zellig S. Harris and Leigh Lisker for valuable comments and suggestions; and to the Editor of Language for his painstaking criticism of two earlier versions of this paper.

² These include dialects of Bolivia (La Paz, Cochabamba), Colombia (Bogotá, Cali), Mexico (Guadalajara), Puerto Rico (San Juan), Nicaragua (Managua, León), and the southwestern United States.

³ Note that in contrast to Castilian, Costa Rican /k/ has no allophone parallel to the voiceless spirant allophones of /p/ and /t/. Castilian /χ/ (a voiceless fricative spirant) and the voiceless non-fricative spirant allophone of /k/ before /s/ both fall together in Costa Rican Spanish with the phoneme /h/.

⁴ Cf. Trager, TCLP 8.217–22 (1939).

⁵ The present treatment of orthographic rr and r as two phonemes /r/ and /r/ (§3.4 and §3.6) is a somewhat radical departure from the usual one, particularly as regards distribution. In Castilian, and probably in most other Spanish dialects, /r/ and /r/ are both trills, and have the following distribution: /r/ initially and medially, /r/ medially and finally. Tho the same analysis for Costa Rican might perhaps be thought desirable in view of the simpler morphology that results from it, it is not supported by the phonetic material of this dialect, and cannot be achieved without invoking extra-phonemic (i.e. morphological) criteria.

In Costa Rican Spanish there is phonetically only one trill [r]; the allophones of /r/ are spirants. The voiceless allophone [r] is in complementary distribution with both [r] and [ř] (the voiced spirant); but phonetically it is clearly nothing but the voiceless correlative of the latter. The distribution of the segments is as follows: voiced [ř] initially and medially; voiceless [r] finally; trilled [r] medially. We have then a phoneme /r/ which occurs in all positions, and another phoneme /r/ which occurs only in medial position. Without invoking data from a later level of analysis, no other description is possible.

⁶ All vowels are slightly nasalized before a nasal consonant; but the nasalization of vowels before /mf/ is much greater and more distinct than before any other nasal or nasal cluster.

⁷ Thus, in desramar ‘to prune’ there is sometimes a slight friction audible before the /r/: fdezřamár]; but just as often the word is phonetically indistinguishable from derramar ‘to spill’: [deřamár].