University of Florida
Pronunciation of Spanish intervocalic /y/112 is subject to considerable regional variation, ranging from a prepalatal affricate to a weakly articulated semivowel . In those dialect regions where the latter pronunciation predominates, total effacement of certain instances of intervocalic /y/ is common, resulting in potential merger of minimal pairs such as sella '(s)he seals' / sea 'be' (subj., 3s.)113. In the same dialect zones, hypercorrect epenthesis of intervocalic /y/ in popular speech is not in frequent: día>diya 'day.' Although these phenomena have been frequently commented, there are few detailed descriptions, and even fewer attempts at incorporating variable elision/epenthesis of /y/ within theories of Spanish phonology. The present study has a dual function: first, to offer a refinement of the observational data, and second, to suggest some tentative steps in the direction of characterizing Spanish dialectal variation in terms of phonological processes. It will be proposed that elision of /y/ results from the Obligatory Contour Principle, operating on an autosegmental tier defining front vowels and /y/. This claim in turn depends upon a model of phonological underspecification of Spanish vowels and glides. The remainder of the discussion proceeds as follows: Section 2 provides data on elision of /y/; Section 3 provides data on epenthesis of /y/; Section 4 discusses underspecification of Spanish vowels and semi-vowels; Section 5 describes elision of /y/ in terms of underspecification and the Obligatory Contour Principle; Section 6 establishes the parallels with hypercorrect insertion of /y/; Section 7 summarizes the discussion.
There is both regional and idiolectal variation as regards the phonetic context of elision of /y/, which together with generally sketchy published observations renders problematic the expression of the basic data. Using data from Mexican-American Spanish, Ross (1980) defines two partially overlapping environments: contact with a preceding or following /i/, and contact with a preceding /e/114:
|patilludo||[patúo]||'having long beard/sideburns'|
|velludo||[beúo]||'hairy (arms, legs, etc.)'|
Similar behavior also occurs in parts of Argentina (Vidal de Battini 1949: 47), El Salvador (Canfield 1960), Nicaragua (Lacayo 1954) and Honduras (Lipski 1987).
In New Mexican Spanish, Espinosa (1930: 197-99) described loss of /y/ in all the environments in (1), plus in /eye/, /aye/ (e. g., calle 'street') and less frequently /oye/ (e. g., oyendo 'hearing'). In Colorado Spanish, Espinosa even found loss of /y/ between two non-front vowels, providing neither was /u/; Henríquez Ureña (1938: 352-53) described similar behavior for parts of Mexico and Guatemala, while Oroz (1966:135-36) found the same patterns of elision in southern Chile.
Finally, there are some Spanish dialects in which /y/, is elided only in contact with /i/, and only rarely if at all in contact with /e/ (e. g., Suárez 1945: 68, Toscano Mateus 1953: 102-03). Given the high degree of variability, particularly in nonstandard varieties, it is difficult to reconstruct the exact route of evolution, but it is likely that elision of /y/ first began following /i/, was later generalized to the mirror-image environment, still later generalized to include elision after /e/, and in a few dialects has generalized to include elision before and after both /e/ and /i/. Elision of /y/ in contact with a following /e/ is frequently impeded when the first vowel is /u/ or /o/, which together with the implicational data just stated suggests that the first vowel exercises greater influence on behavior of intervocalic /y/. In contrast to the previously mentioned cases, elision of /y/ between two non-front vowels is much more sporadic, is limited to a few lexical items in each dialect, and by all evidence is a manifestation of general deletion of intervocalic consonants, rather than being tied to the quality of the surrounding vowels. In the following analysis, which will not refer specifically to individual dialects, elision of intervocalic /y/ will be assumed in contact with a preceding or following /i/ or /e/, this being the obvious generalization. Dialects exhibiting other patterns can be included through appropriate modifications to the theoretical model to be developed below, which is not crucially dependent on interdialectal variation.
Frequent in weakly monitored/nonstandard speech of Spanish dialects characterized by elision of intervocalic /y/ is epenthetic insertion of /y/ in hiatus combinations which precisely duplicate environments for elision of /y/ (Flórez et. al. 1969: 93; Toscano Mateus 1953: 103; Canfield 1960, Rabanales 1950, García Fajardo 1984: 51-52, etc.):
|caer > cayer||'to fall'|
|leer > leyer||'to read'|
|día > diya||'day'|
|vea > veya||'look (imp.)'|
|maíz > mayiz||'corn'|
|viuda > viyuda||'widow'|
|tío > tiyo||'uncle'|
For a subset of speakers in the dialects in question, elision and epenthesis have progressed to the point where it might appear that /y/ is no longer in the underlying representation of the words in question, so that, for instance, leyes 'laws' and lees 'you (fam). read' are now homophones at the underlying level115. For most speakers, however, this extreme restructuring is approached only asymptotically, but the symmetry with respect to elision to /y/ remains.
The status of Spanish glides which alternate with consonants (e. g., ley [l] 'law' vs. leyes [leyes] 'laws') has been the subject of considerable theoretical discussion116. In contrast, the status of morpheme-internal intervocalic /y/ has inspired a greater consensus, with the major points of discord being the alternative specification of this element as an underlying consonant [+ consonantal] or as an underlying (nonsyllabic) [+vocalic] element which is subsequently 'promoted' to full consonantal status when onset-initial.
In Spanish dialects where elision of intervocalic /y/ does not occur, the resulting segment invariably has consonantal characteristics, ranging from fricative to affricate articulation, and it is feasible to postulate an underlying consonant (cf. Harris 1983: 61), specified [+ consonantal] or an equivalent designation, and corresponding to a C slot on the skeleton. On the other hand, dialects such as those of Central America present an intervocalic segment which is more transparently a semivowel . Theories of phonological underspecification offer new ways of viewing the relationship between full vowels and semi vowels/glides. In particular, the distribution of, e. g., Central American Spanish intervocalic /y/ vis-à-vis the full vowel /i/ can be accounted for by underspecification of the feature [syllabic]. What emerges as intervocalic /y/ in these dialects is underlyingly unspecified for syllabicity, thus corresponding to a C slot on the skeleton (e. g., Steriade 1984), but specified [-consonantal] (cf. also Hyman 1985: chap. 6; Guerssel 1986, who, however, postulate no underlying values for syllabicity). Syllabicity (in this case [-syllabic]) is supplied by a process such as Steriade's (1984) 'CV rule'. The remaining features associated with /y/ are those which define vowels.
Assuming a model of radical underspecification of the
Spanish vowel system (cf. Archangeli 1984, 1988), it is possible to have one
An analysis of /y/-dropping as a result of the Obligatory Contour Principle (OCP), to be motivated shortly, requires that the feature value [-back] form a separate tier at some point in the derivation. In models in which /e/ is the maximally underspecified vowel, /e/ and /i/ are not underlyingly specified for frontness; this value is supplied by a redundancy rule. Bearing this in mind, the subsequent analysis can take two different directions. The first involves choosing an alternate set of underlying specifications in which /i/ and /e/ are specified [-back] (e. g., the model of Vago 1988 in which /a/ is the default vowel). The second allows for operation of the OCP on a tier defined by values which are not present underlyingly, but rather introduced by feature-filling rules. The evidence in favor of default status for Spanish /e/ is compelling, while there is much justification for the intermingling of feature-filling redundancy rules and phonological rules (cf. Archangeli 1984). Therefore I adopt as a working hypothesis the application of the OCP on tiers representing feature values not present underlyingly, together with the following underlying feature specifications and rules for Spanish vowels:119
The behavior of intervocalic /y/ in Spanish suggests an intimate connection between the phonological specification of /y/ and the specification of the adjacent vowels. As noted in Section 2, there are two fundamentally different processes resulting in elision of intervocalic /y/. The first process, represented by only a few extreme cases, potentially involves loss of all instances of intervocalic /y/, including between two low vowels (calla '(s)he is quiet'), between two non-front vowels (caballo 'horse'), and so forth. At work here is (variable, lexically constrained and idiosyncratic elision operative irrespective of the surrounding vowels. Such a process has no immediate consequences for underspecification theory, and will not be discussed further120.
The second process, of central concern to the present endeavor, involves loss of /y/ in contact with front vowels. The minimal feature specification [front] is involved in these cases121, so that one could postulate that in contact with a [front] vowel the (maximally underspecified) feature matrix underlying /y/ receives the value [front] through autosegmental spreading from one of the contiguous vowels. While formally possible, this approach has drawbacks which are not outweighed by possible advantages. First, it requires that superficial [y] be given two fundamentally different derivations depending upon the surrounding environment, an undesirable consequence unless independent justification can be offered. Assuming default left-to-right spreading, such an approach gives no ready explanation for putative spreading of [front] from a FOLLOWING vowel (as in gallina). More importantly, there is no principled explanation of why only [front], and not other features, spreads to the empty matrix. In the absence of evidence to the contrary, I therefore assume that the [front] specification of intervocalic /y/ results from application of (3b), while the eventual [high] value arises through the application of a later rule.
At the root of elision of intervocalic /y/ is the
juxtaposition of phonological matrices containing the [front] specification.
This pattern of behavior is the signature of the Obligatory
Since elision of Spanish intervocalic /y/ occurs in contact with /i/ and /e/, it is clear that complete identity of phonological matrices is not required, but only a [front] specification. For these dialects, [front] forms a separate tier, on which OCP effects manifest themselves. In those instances where /y/ occurs in contact with a [front] vowel, OCP violation is avoided through delinking one of the slots from the [front] tier and spreading [front] from one of the neighboring vowels, creating a multiply-finked structure. This is shown schematically in (4) in the case where the first vowel is [front], but the rule also operates in the mirror-image environment:
It might be postulated that elision of /y/ is a single-stage process, namely delinking not followed by spreading, resulting in eventual deletion of the corresponding skeletal slot. However, words affected by (4) interact with low level resyllabification rules, as well as with fast-speech rules raising unstressed /e/ and /o/ in hiatus: teatro [tiá-tro] 'theater,' poesía [pue-sí-a] 'poetry' (Navarro Tomás 1967: 68 f.). At times, /y/-deletion appears to operate after the processes just described. Thus, it is relatively uncommon for an unstressed [i] brought into hiatus through elision of /y/ to form a new diphthong gallinazo 'parasitic plant' is usually [ga-i-ná-so) rather than [ga-ná-so]. Sellado 'sealed' is routinely pronounced as [se-á-do], but very seldom as [sí-á-do] and virtually never as *[sádo]. Similarly, the hiatus created through elision of /y/ in bi(ll)ete 'bill' is rarely resyllabified as a diphthong (cf. siete 'seven'). However, exceptions to some of the above-mentioned restrictions occasionally arise in rapid speech, suggesting that low-level rules take as their input structures to which (4) has already applied. Since tier conflation, after which a [front] tier is no longer available to trigger OCP effects on /y/, occurs before the application of fast speech rules, this variability requires explanation. I propose that it is precisely the linking of multiple skeletal slots to a single autosegment, as in (4), which in conjunction with other phonological processes in Spanish, provides the basis for a principled explanation.
Assuming that the [high] specification of /y/ is added by a
relatively late application of 'glide raising',122 then
at the point of application of the OCR, /y/ is only specified as [front].
Detachment of this feature in effect detaches the entire Dorsal articulator
node. The slot corresponding to /y/ is reattached to the [front] tier via a
neighboring vowel, but acquires no other feature specifications at this
As a consequence of this portrayal, elision of /y/ results from the concerted action of TWO mechanisms: the creation of multiply-linked structures via rule (4), resulting from application of the OCP early in the derivation, and phonetic implementation rules which at times merge /y/ with an adjacent vowel. Creation of new diphthongs from instances of unstressed [i] brought into hiatus through elision of /y/ can in principle be impeded, since the slot associated with /y/ remains throughout the derivation, although implementation rules may deprive it of phonetic substance.
Insertion of /y/ to break certain hiatuses is not a priori tied to loss of intervocalic /y/ (particularly since insertion of other hiatus breaking consonants is frequent in the history of Spanish; cf. Malkiel 1958), but the existence of a correlation between the two phenomena is supported by two powerful bits of circumstantial evidence. First, epenthetic /y/ occurs in the same dialect zones in which intervocalic /y/ is subject to loss. Second, epenthesis of /y/ occurs in precisely the same vocalic environments that condition loss of /y/.123 The appropriate account of /y/-epenthesis is adjunction of an unspecified (vocalic) skeletal slot in the proper environment (between two vowels, at least one of which is [front], followed by application of the redundancy rules in (3). The resulting combination then incurs the effects of the OCR, including application of (4), thereby manifesting the same range of phonetic realizations as underlying /VyV/. This is shown in (5), where % symbolizes a mirror-image environment:
The potential reversibility of epenthesis and elision of /y/ underscores the fact that in this group of Spanish dialects, the feature [front] forms a separate tier which governs the superficially variable appearance of /y/.
The following claims have emerged from the preceding discussion:
(a) In Spanish dialects characterized by elision of intervocalic /y/, this segment is not underlyingly specified as C, but rather as a minimally specified V.
(b) the eventual [front] specification of /y/ arises through a feature-filling rule defining the default vowel, while [high] comes about through a subsequent rule which raises non-syllabic vowels.
(c) In contact with another [front] segment /i/ or /e/, operation of the Obligatory Category Principle on a [front] tier causes delinking of /y/ and linking to the [front] associated with the adjacent vowel.
(d) /y/ thus dually linked cannot acquire [high] through glide raising, and emerges from the phonological component with only a [front] specification. Phonetic implementation rules account for variable realization of this element, ranging from a full semivowel to total elision.
(e) Epenthesis of /y/ involves adjunction of an empty slot between a [front] vowel and another vowel, followed by application of feature-filling rules and subsequent OCP effects.
In summary, Spanish dialects in which /y/ dropping is
frequent contain a maximally underspecified (and eventually nonsyllabic)
vocalic slot instead of a consonantal slot underlying intervocalic /y/. This is
in effect the 'parameter' separating /y/-dropping dialects from other Spanish
dialects. Among those dialects in which /y/ emerges as a semivowel, elision in
contact with front vowels is the usual
The analysis just presented lends support to expanded interpretation of the OCP as operative on tiers defined by individual features, and to radical underspecification of the Spanish vocalic system. It also offers insight into the frequent loss/epenthesis of /y/ in contact with front vowels, phenomena which occur in many languages.
Alonso, Amado. 1967. Estudios lingüísticos, temas hispanoamericanos. Madrid: Gredos, 3rd ed.
Alvar, Manuel. 1977. Leticia: estudios lingüísticos sobre la Amazonia colombiana. Bogotá: Instituto Caro y Cuervo.
Archangeli, Diana. 1984. Underspecification in Yawelmani Phonology and Morphology. New York: Garland Publishing, 1988.
_____.1988. «Aspects of Underspecification Theory». Phonology 5: 183-207.
Barrutia, Richard and Tracy Terrell. 1982. Fonética y fonología españolas. New York: Wiley.
Borzone de Manrique, Ana, 1976. «Acoustic Study of /i, u/ in the Spanish Diphthong». Language and Speech 19: 121-28.
Bowen, J. Donald. 1974. «New Mexican Spanish Verb Forms». Garland Bills (ed.), Southwest Area Linguistics, 161-62. San Diego: University of California.
Canfield, D. Lincoln. 1960. «Observaciones sobre el español salvadoreño». Filología 6: 29-76.
_____. 1981. Spanish Pronunciation in the Americas. Chicago: University of Chicago.
Clements, George N. 1985. «The Geometry of Phonological Features». Phonology Yearbook 2: 223-50.
Cressey, William, 1978. Spanish Phonology and Morphology: A Generative View. Washington: Georgetown University.
Espinosa, Aurelio. 1930. Estudios sobre el español de Nuevo Méjico; parte I: fonética. Trans. Amado Alonso and Ángel Rosenblat. Buenos Aires: Biblioteca de Dialectología Hispanoamericana 1: 19-313.
Flórez, Luis, José Joaquín Montes Gualdo and Jennie Figueroa Lorza. 1969. El español hablado en el departamento del Norte de Santander: Bogotá: Instituto Caro y Cuervo.
García Fajardo, Josefina. 1984. Fonetica del español de Valladolid, Yucatán. México: Universidad Nacional Autónoma de México.
Guerssel, Mohamed. 1986. «Glides in Berber and Syllabicity». Linguistic Inquiry17: 1-12.
Halle, Morris. 1988. «The Immanent Form of Phonemes». William Hirst (ed.), The Making of Cognitive Science, Essays in Honor of George A. Miller, 167-83. Cambridge: Cambridge University Press.
Harris, James. 1969. Spanish Phonology. Cambridge: M. I. T. Press.
_____. 1970. «Sequences of Vowels in Spanish». Linguistic Inquiry 1: 129-34.
_____. 1983. Syllabic Structure and Stress in Spanish. Cambridge: M. I. T. Press.
_____. 1985. «Spanish Word Markers» Frank Nuessel, Jr. (ed), Current Issues in Hispanic Phonology and Morphology, 34-54. Bloomington: Indiana University Linguistics Club.
_____. 1987 «Epenthesis Processes in Spanish». Carol Neidle, Rafael Núñez Cedeño (eds.), Studies in Romance Languages, 107-22. Dordrecht: Foris.
Hayes, Bruce. 1986. «Assimilation as Spreading in Toba Batak». Linguistic Inquiry 17: 467-99.
Henríquez Ureña, Pedro. 1938. El español en Méjico, los Estados Unidos y la América Central. Buenos Aires: Biblioteca de Dialectología Hispanoamericana 4.
Hutchinson, Susan 1974, «Spanish Vowel Sandhi». Papers from the Possession on Natural Phonology 184-91. Chicago: Chicago Linguistic Society.
Hyman, Larry. 1985. A Theory of Phonological Weight. Dordrecht: Foris.
Lacayo, Heberto. 1954. «Apuntes sobre la pronunciación del español en Nicaragua». Hispania 37: 267-68.
Ladefoged, Peter. 1988. «Hierarchical Rules of the International Phonetic Alphabet» BLS 14: 124-41.
Lipski, John. 1987. Fonética y fonología del español de Honduras. Tegucigalpa: Editorial Guaymuras.
_____. 1989. «Theoretical Consequences of Ecuadoran Vowel Reduction» Hispanic Linguistics, forthcoming."
Malkiel, Yakov. 1958. «Los interfijos hispánicos». Diego Catalán (ed.), Miscelánea homenaje a André Martinet, t. II, 107-99. La Laguna. Universidad de La Laguna
McCarthy, John. 1986. «OCP Effects: Gemination and Antigemination». Linguistic Inquiry 17: 207-63.
Navarro Tomás, Tomás. 1967. Manual de pronunciación española. New York: Hafner, 6th ed.
Núñez Cedeño, Rafael. 1986. «La /s/ ultracorrectiva en dominicano y la estructura silábica». José Moreno de Alba (ed.), Actas del II congreso internacional sobre el español de América, 337-47. Mexico: Universidad Nacional Autónoma de México.
_____. 1988. «Structure-preserving Properties of an Epenthetic Rule in Spanish». David Birdsong, Jean-Pierre Montreuil (eds.), Advances in Romance Linguistics, 319-35. Dordrecht: Foris.
Odden, David. 1986. «On the Role of the Obligatory Contour Principle in Phonological Theory». Language 62: 353-83.
Oroz, Rodolfo. 1966. La lengua castellana en Chile. Santiago: Universidad de Chile.
Peñalosa, Fernando. 1980. Chicano Sociolinguistics. Rowley, Mass.: Newbury House.
Predmore, Richard. 1945. «La pronunciación de varias consonantes en el español de Guatemala». Nueva Revista de Filología Hispánica, 7: 277-80.
Pulleyblank, Douglas. 1986. Tone in Lexical Phonology. Dordrecht: Reidel.
Quillis, Antonio. 1981. Fonética acústica de la lengua española. Madrid: Gredos.
Rabanales, Ambrosio. 1950. «Hiato y antihiato en el español vulgar de Chile». Boletín de Filología 12: 197-223.
Ross, J. Ronald. 1980. «La supresión de /y/ en el español chicano» Hispania 63: 552-54.
Sagey, Elizabeth. 1986. «The Representation of Features and Relations in Non-linear Phonology». Diss. M. I. T.
Sánchez, Rosaura. 1972. «Nuestra circunstancia lingüística». El Grito 6 (1): 45-74.
Schein, Barry and Dona Steriade. 1986. «On Geminates». Linguistic Inquiry 17: 691-744.
Steriade, Dona. 1984. «Glides and vowels in Romanian». BLS 10: 47-64.
_____.1987. «Redundant Values». CLS 23/2: 339-62.
Suárez, Víctor.1945. El español que se habla en Yucatán. Mérida: Díaz Massa.
Terrell, Tracy. 1982. «Relexificación en el español dominicano: implicaciones para la educación». Orlando Alba (ed.), El español del Caribe, 301-18. Santiago de los Caballeros: Universidad Católica Madre y Maestra.
_____. 1986. «La desaparición de /s/ posnuclear a nivel léxico en el habla dominicana». Rafael Núñez Cedeño, Iraset Páez Urdaneta, Jorge Guitart (eds.), Estudios sobre el español del Caribe, 117-34. Caracas: Casa de Bello.
Toscano Mateus, Humberto. 1953. El español del Ecuador. Madrid: Consejo Superior de Investigación Científica.
Wilson, Jack. 1971. «A Generative Phonological Study of Costa Rican Spanish». Diss. University of Michigan.
Vago, Robert. 1988. «Underspecification in the Height Harmony System of Pasiego». Phonology 5: 343-62.
Vidal de Battini, Berta 1949. El habla rural de San Luis. Buenos Aires: Instituto de Filología.
Yip, Moira.1988. «The Obligatory Contour Principle and Phonological Rules: A Loss of Identity». Linguistic Inquiry 19: 65-100.
Zamora, Juan Clemente and Jorge Guitart. 1988. Dialectología hispanoamericana. Salamanca: Publicaciones del Colegio de España, 2nd ed.