LIQUID ASSIMILATION IN HAVANA SPANISH

There has been a consensus among linguists that laryngeal and superlaryngeal nodes are located under the root node and place node is under the superlaryngeal node. However, there is very little consensus on where manner of articulation features belong. A phonological analysis of the spreading and delinking process occurring in the educated Spanish of Havana (Cuba) suggests that: (1) the continuant feature should be located under the superlaryngeal node; (2) continuant and place features form a natural class node (oral cavity); and (3) manner features and place features do operate together in the delinking and spreading operations. (Author/MSE) *********************************************************************** Reproductions supplied by EDRS are the best that can be made from the original document. *********************************************************************** "PERMISSION TO REPRODUCE THIS MATERIAL HAS BEEN GRANTED BY TO THE EDUCATIONAL RESOURCES tNFORMATION CENTER (ERIC)." U.S. DEPARTMENT OF EDUCATION Office of Educational Research and Improvement EDUCATIONAL RESOURCES INFORMATION CENTEP (ERIC) document hes been reproduced as received from the person or organization originating it O Minot changes have been made to improve reproduction quality Points of view or opinions stalso in this docu . mint do not necessarily represent official OER1 position or policy Cr..1 LIQUID ASSIMILATION IN HAVANA SPANISH Fenfang Hwu AbAtriket: There has been a consensus that Laryngeal and Supralaryngeal nodes are located under the Root Bode and Place node I. under the Supralaryngeal node (Clements 1985, Sagey 1987, Archangeli and Pulleyblank 1988). There is, however, very little consensus on where manner of articulation features belong. It will be argued in this paper, basing on spreading and delinking process occurring in Liquid assimilation in Havana Spanish, that 1) the [continuant] feature should be located under the Supralaryngeal node 2) [continuant] and place features form a natural class node: Oral Cavity as proposed in Clements (1987) 3) manner features and place features do operate tc.gether in the delinking and spreading operations. According to Cuitart (1978, 1980) and Zamora & Guitart (1982), in the educated Spanish of Havana, the distinction between Ill and Ir] in syllable final position is lost. The following data show this phenomenon. (1) (Guitart 1976, 1980; Zamora & Guitart 1982; Harris 1985) a.L + Labial ser bobo, ellobo serpobre, eLpobre ser mata, tal mats Beano, eljno b.L Coronal arde, falda ser te, eLW ser nata, taLtata farsa, falsa combo, cokildn ser rojo, eojo Carlos, estar lejos ser droga, tal droga (1)14 [bp] [min] BEST COPY AVAILABLE 2 Kansas Working Papers in Linguistics, 1991, Volume 16, pp. 59-75

a natural class node: Oral Cavity as proposed in Clements  (1987) 3) manner features and place features do operate tc.gether in the delinking and spreading operations.
According to Cuitart (1978, 1980) and Zamora & Guitart (1982), in the educated Spanish of Havana, the distinction between Ill and Ir] in syllable final position is lost.The following data show this phenomenon.The .kbove data show that liquids assimilate to the following segments in some way.Retroflection occurs only when the triggering segment of assimilation is corona'.However, when the liquid is followed by a consonant cluster or palatal the retroflection does not occur.With the triggering segment being labial or palatal, we get geminates (mm] and [ill] but not when the triggering segment is an alveolar nasal: seldnlata.From the above observations we conclude that corona's except palatais behave asymmetrically with respect to the other consonants. It has recently been argued by Kiparsky (1985) and Avery & Rice (1988, 1989) among others, that, beside being more frequent, corona's are more prone to undergo assimilation processes than any other place of articulation.Puel, Nespoulous, Bonafd and Rascol (1980) point out that coronals also pattern in a unique way in aphasic speech, thus providing external evidence in favor of the claim that coronals have a special status.Avery & Rice (1989) have claimed that the special status of corona's lies in the fact that they lack a specification for place features in 'underlying representation.According to them the behavior of corona's in Catalan presents one evidence for the underspecification of the Corona!node.
(2) (Kiparszy 1885) a. /n/ The above data show that the underlying alveolar /n/ assimilates to the point of articulation of the following segment.But the underlying labial /m/ assimilates only within its own primary content node, naLaely labial and labiodental.According to Rice & Avery (1289) the phonology is restricted to at most three operations: spreading, delinking, and OCP- based fusion.Spreading is considered as-a language-particular operation which may clude trigger and target conditiola as well u a directionality parameter.The theory of spreading that they adopt is that (3) a. spreading cui occur only if a structural target is present, and b. a feature or node spread only to an empty position.
Fusion is an operation which takes identical primary content nodes and fuses them provided that the nodes are non-distinct; i.e. both nodes do not dominate different secondary nodes.Fusion is headed in that the secondary features of the triggering segment are maintained.Normally it takes place under conditions of strict adjacency.Delinking is a neutratzation process which delinks content nodes in neutralizing positions such as morpheme-final or syllable-final.
Basing on the notions above, the Catalan case is analyzed (Avery & Rice 1089) as follows.Since coronals are considered as underspecieed, two distinct phonological processes are proposed.One process resulting in nasal assimilation is spreading of the place features of the non-nasal segment to an empty Place node.This accounts for the alveolar nasal- labial and alveolar nasal-velar clusters as shown in (4).The ot1-2 process is fusion.Since Lie Coronal node is absent underlyingly, when a redundancy rule applies filling in Coronal specification, the structural description of fusion is met and it must automatically apply.This 4 accounts for the alveolar nasalcoronal sequences (5a).Fusion also account for the forms with labial nasallabial sequences, but it takes place before the redundancy rule since labials are specified for Labial node underlyingly (54 (4) (5) a.In standard Spanish, the alveolar nasal /n/ behaves in the same way as in Catalan.However, the labial nasal /m/ and the palatal nasal /ii/ do not.In Spanish alve)lar nasal /n/, labial nasal /m/ and palatal nasal /11/ cord,rast In syllable iaitial position.In syllable final position only alveolar nasal /n/ occurs underlyingly and it assimilates to the place of articulation of the following segment (6).As shown in (7) another segment in Standard Spanish that shows a similar type of assimilation is alveolar lateral /1/.It assimilates to the following segment when the following segment is corona).Since coronals are claimed to be underspecified when analyzing the Catalan nasal assimilation (Avery & Rice (1989)) the same analysis can be adopted to explain the nasal assimilation process in spanish.However, lateral assimilation in Standard Spanish does not follow the same analysis.
If we claimed, following Avery & Rice (1989), that the lateml /1/ was underspecified, we would get laterals assimilated to every following segment.Therefore the only explanation for laterals not assimilating to labials or velars is that laterals are specified for Coronal node underlyingly, therefore the labial or velar features cannot be spread onto it.Only the fusion rule that fuses identical primary content nodes can account for the assimilation.Some current research assumes that (lateral] is a coronal dependent because laterals are almost alwaya coronds (Levin (1988), McCarthy (1988)).However, Rice and Avery, and Shaw (1991) argue that, while it is true that laterals are normally coronal, it does not follow that [lateral] is a coronal dependent.In this analysis lateral cannot be considered as a coronal dependent since the fusion rule will make sure that lateral share the same place node as the following segment, which would result in losing [lateral] feature if (lateral] was located under the Corona!node.Therefore, we conclude that In standard Spanish alveolar nasal /n/ is underspecified for Corona) node underlyingly and alveolar lateral /1/ is 6 specified for Coronal node, which means that the default coronal specification applies very early in the k.xicon.
In Havuna Spanish /n/ and /1/ assimilate but not in exactly the same way as in standard spanish.In Havana Spanish every nasal in coda, regardless of the consonsnt that follows, is normally velarized, although assimilation also takes place in some cases, resulting in a sort of co-articulation where a dorsovelar element is always present (Hattis 1969).The velar nasal occurs before dental and alveolar and alveolo-palatal consonants, and, of course, before velars, while a velarized iabial nasal occurs before labial and labiodental consonants.In utterance-final position, however, it is only the velar nasal that occurs.This operation call be explained if we consider coronals as underspecified.First, velarization changes all syllable final nasals into velar nasals (9).Then spreading will spread the place features from the following segment onto nasal creating a complex segment (10 a & Ve I Since coronals are still underspecified for Corona) node when the spreading takes place, they can not spread any features onto the preceding velarised nasal (11).Therefore, we get a phrase like 411 domingo.When redundancy rule applies filling in the Coronal node to the coronal segment, the nasal-coronal sequence cannot fuse because they do not have the same primary place node.# Fits:on -> coR However for this analysis to work we have to relax the second condition of Spreading (3) in such a way that the target of the spreading can be non- empty when spreading will create a complex segment.The spreading of the dorsal feature to the preceding nasal will create a geminate velar nasal, which will be simplified into one velar (lob).Therefore, we get a phrase like uilj caballo.With respect to the alveolar lateral /1/ the same type of assimilation as in standard Spanish occurs except that palatal does not trigger assimilation.Based on the above analysis we might conclude that except lateral, corona's are underspecified in Havana Spanish.However, palatals which are considend as corona's do not pattern with corona's either.It has been suggested by Keating (1987) In this analysis we will assume that corona's, except lateral /1/, lack primary content node underlyingly as reprmented in (12) and will be filled in by redundancy rule at a later stage of derivation.We also adopt the notions of spreading, delinking and fusion proposed by Avery & Rice (1989).PiAg.c corolui Since in absolute hal position liquids either maintain their identity or are realised as a retroflex (s).Harris (1985) proposes a retroflection rule which changes syllable-hal liquids int4 a retroflex @I.We will adopt the same rule as shown in (13).
(13) a. abril= &brill] abrir= abri [r] or abrifsl] b.Rhyme Retroflection (Harris 1985 Mter the Retroflection the second process is that the retroflex segment assimilates to the following segment in place and manner of articulations.This process will involve delinking and spreading.Iverson  (1989) proposes the elimination of Supra laryngeal node.This process of assimilation then shows that features under Supra laryngeal node do operate together in a phonological operation, namely delinking and spreading ( 14).prominently in the work of Clements (1985,1987), Sagey (1985) and However, this process does not apply to "L + coronal" group.This will be attributed to the absence of Coronal node underlyingly.Hence we suggest that the spreading, besides what have been suggested by Avery and Rice (1989), be restricted by the following constraint: the spreading can occur only when the trigger is specified with the primary content node if it involves place assimilation.
The reason for [v] turning into (fl (15) is that iy] almost never occurs in Havana Spanish (Haden & Matluck 1973).Hence we will need a low level rule to derive the right result.

Low level rule: (+lab, +coat, +stri] [-voice]
The glottal sound /h/, as is well known, carries with it no specific place of articulation features, but rather adopts that of neighboring vowel.
We will need a low level rule which turns a segment with vowel place In the following we will present rules that derive "L + coronal" group.These rules are ordered as follows: a. redundancy rule that fills in the primary content node with Coronal, b. fusion, which requires strict adjacency and takes the left segment as the hnd, and c. regressive assimilation which takes the Oral Cavity node as a group for assimilation.
Liquid will turn into retroflex [4] before the following corpnal segment being filled in with Coronal by redundancy rule, resulting in the context for flision.The fusion will be headed to maintain the secondary feature of the left segment.The regressive assimilation then takes place assimilating the left segment, which already shares the same place features 83 the right segment, to the right segment in place of articulation and continuant 12 feature.This assimilation supports Clements' proposal that Oral Cavity, which dominates Place node and continuant features, is a natural class node.On the other hand it also provides evidence, beside Davis (1989), that continuant is located under the Supra laryngeal node.The derivations of the word farsa 13 shown in the following.The derivations of the phrase ser nata is shown in the following.

Ant
As shown in (1b) the retrofiection does not occur when a consonant cluster follows the liquid.This is because that fusion takes syllable structure into account.Therefore the two segments in onset cluster for example /t/ and /r/ will fuse first resulting in a multiply linked structure as shown in ( 20).If we follow Schein & Steriade's claun (1980) that "if the structural description of some rule affecting a imposes conditions that are met by # but not by i then the rule will not apply to a multiply linked structure" like (21) and that linear precedence is defined between root nodes or between skeletal units and we restrict the fusion to take place under strict adjacency, the reason that the fusion does not take place in the above described situation is explained.The derivations of the phrase ser fres is shown in ( 22).
1.4 In this analysis we have seen that Supralaryngeal node is operative in the phonological process and place features and continuant feature form a natural group and by considering coronals as underspecified we achieve to explain their special behavior. 15 (6) (Hualde 1989) co[m] piedra.sco[9] fuersaco[n] cera 'with stones' 'with strength' 'with wax' supports the existence of Supralaryngeal category advocated most Archangeli & Pulleyblank (forthcoming).It also supports the suggestion that the continuant feature is located under the Supralaryngeal node (15 a. features and 1i-continuant) into a lb].Low level rule: [+cont, vowel place] -4+voice] order to derive the correct result we need a low level rule:[+cont, +voice, -anti +(+son] Lipski (1989)latograms that palatals actually have two articulator nodes: Coronal and Dorsal.Lipski (1989)presents distributional and articulatory data to support that in Spanish palatals 10(/ and /fi/ are complex segments i.e., having both Coror al and Dorsal nodes.According to him, pram) ((eq--,[y]) that is sometimes described as "delateralisation" and /fi/-gliding can be represented as delinking of the Coronal articulation node.This then can explain why and 61 behave differently with respect to the other coronals in the Liquid Assimilation of Havana Spanish (1c).That is because that carries only Dorsal node and that [fl] carries both Cormal and Dorsal nodes and the presence of the Dorsal node forcc.i the presence of Coronal node underlyingly.
In this analysis we assume that gi is a contour segment.It contains two Oral Cavity nodes.Each one carries one continuant feature.However, the regressive assimilation will only spread the cloeer (left) Oral Cavity node.