On the Specification of Coronals

ON THE SPECIFICATION OF CORONALS Blanca Palmada and Pep Serra Universitat Autbnoma de Barcelona We present the relation between assimilatory asyrnmetries in Central Catalan and the underspecified character of Coronal. We account for differences in the assimilatory behaviour of coronals by placing Lateral and Strident under Coronal. We assume that these features are dependents of Coronal and that this justifies the presence of the unmarked articulator in the underlying representations of sirident and lateral coronals. The assimilatory possibilities of centralised consonants in Alguerese Catalan force us to argue that Lateral and Strident are double-linked features. We consider an altemative analysis based on Configurational Constraints and we demonstrate that this analysis fails to capture the gliding processes which affect some complex segments in Majorcan Catalan and Occitan. The organisation of features allows us to provide a simple and unitary account, in this way the enrichrnent of the representations derives in a simplification de component. Universal Grammar provides a markedness theory which supplies inforrnation as to which features are underspecifiable. In ths way, the unmarked articulator node seems to be the Coronal one and thus it must be absent in underlying representations. The purpose of this paper is to provide evidence in support of Underspecification theory and to use thls model to investigate the way a feature predicted to be absent in underlying representations can be active in the phonology of a language. In line with current phonological research, our proposa1 is consistent with the claim that the burden of explanation should be in the representational component rather than in the rule Catalan Working Papers in Linguistics (CWPL), 1991 : 181 - 199 Universitat Autbnoma de Barcelona 1. Assimilatory Patterns in Central Catalan It has been observed that coronals behave asymmetrically with respect to other consonants in some languages and this is the case of the coronal nasal and the stop in Catalan, which assimilate to all places of articulation. It is generally assumed that this behaviour derives from the general theory of Underspecification. The asyrnmetric pattern of coronals can be illustrated by the data in (1) which correspond to an optional process of stop assimilation in Central Catalan. While the coronal stop assimilates to all places of articulation, as can be seen in (la), labial and velar stops in (lb) and (lc) assimilate 182 component. Our argumentation is based on processes of assimilation, centralisation and gliding which can be found in Catalan and Occitan. This paper is organised as follows. In Section 1 we present the relation between assimilatory asymmetries in Central Catalan and the underspecifíed character of Coronal and we hypothesize that these asymmetries have their origin in the Feature Geometry. We conclude that Lateral and Strident are dependents of Coronal and that this justifies the presence of the unmarked articulator in the underlying representations of lateral and strident coronals. We also sketch two earlier analyses of the way Coronal is introduced in representations: Avery-&ce (1989) and Paradis-Prunet (1989). We argue that such analyses cannot solve the problem we present. Section 2 deals with data coming from Alguerese Catalan. We argue from the assimilatory possibilities of centralised consonants that Lateral and Strident are double-linked features. We consider an alternative analysis based on Configurational Constraints and in Section 3 we demonstrate that this analysis fails to capture the gliding processes that affect some complex segments in Majorcan Catalan and Occitan. We conclude that the processes considered must be treated in a unified manner and not as accidents. The organisation of features allows us to provide a simple and unitary account. In this way we follow the usual strategy to obtain that the enrichment of the representations derives in a simplification rule component. only within their primary content node. This is exactly the behaviour that we predict by underspecifyng the Coronal node. (1) a. se [tl se [m] mans se [P,] focs se [I] línies se [t, S]ais se [Llilibres se [k] cases se [d] dones se [b bleus C. po [kl po [kl pa po Ikl foc po [kl sol po [k Slai po [k ai bre po [kl cas 'seven' 'seven hands' 'seven fires' 'seven lines' 'seven lambs' 'seven books' 'seven houses' 'seven women' 'seven voices' 'few' 'few bread' 'few fire' 'few sun' 'few lamb' 'few book' 'few case' b. ca [PI 'no' ca [m] mB 'no hand' ca [P,] foc 'no fire' ca [p] signe 'no sign' [P $]A 'no lamb' ca [p Llibre 'no book' ca [PI casa 'no house' Following the strategy that phonological processes are best explained through a focus on the representational component, we view the assimilation process as a simple operation of spreading, which takes place from right to left and only to an empty position. To make this clear, compare (2a) and (2b). The Dorsal content node spreads to the empty Place node in (2a), but spreading is blocked in (2b) because the Place node of the first consonant has the dependent Labial node and we don't allow cases of spreading triggering delinking. (Note that, strictly speaking, the SL and Place nodes in the representation of [t] are not present underlyingly). (2) a. b. [t I [kl [P 1 [kl X X X IrR 1 Place 1 Pke lke "SL. Place .L \ % -.. --. -k -'+- 'X, %% --\ --% Dorsal Lablal Dorsal [b=kl [md I [bxkl se [t] [klases -> se [kk] ases ca [p] Blasa -> ca [pklasa 'seven houses' 'no house' Consider now the nasal assimilation data. The coronal nasal, as does the coronal stop, patterns as if it has no specified Place of articulation (see the examples in 3a). The other nasals assimilate only within their own primary content nodes: the labial in (3b) can assimilate to a labiodental, and the velar and palatal nasals in (3c) do not assimilate at all. (3) a. s6 [n] 'they aret s6 [m] pocs 'they are few' s6 [q] feliqos 'they are happy' s6 [n] dos 'they are two' s6 [n] sincers 'they are sincere' s6 [n] rics 'they are rich' s6 [n, 3lermans 'they are brothers' s6 [n, Lliures 'they are free' s6 [IJ] grans 'they are big' C. ti 191 Pa 'I have bread' a [F] felig 'happy year' b. so [ml so [ml pocs so [rlj] feli~os so [m] dos so [m] sincers so [m] rics so [m 3lermans so [m oiures so [m] grans 'we aret 'we are few' The assimilation of the coronal nasal is parallel to that of the coronal stop. Its articulation is totally dependent on the following consonant. We predict this behaviour by underspecifing the coronal node in the representation of the nasal in (4). If Coronal is the unmarked specification for Place, then we expect to find a unique patterning of Coronals at least in relation to assimilatory processes. But this is not what we find if we look at the behaviour of the lateral and fricative coronals. They pattern as the other consonants do and their assimilatory possibilities are restricted to their own primary content node. We will demonstrate that Coronal is present in the underlying representations of [I] and [s] and that thls may be used to explain their assimilatory pattern. Lateral assimilation differs from the assimilation of stops and nasals in that laterals assimilate only if they can preserve their coronal character, as can be seen in (5). (5) e @I e [ll pa e @I foc e @] dia e [l] sol e [I] ric e U, 31ema e [l, Llibre e 111 gos 'the' 'the bread' 'the fire' 'the day' 'the sun' 'the rich' 'the brother' 'the book' 'the dog' (Mascaró (1976)) In the analysis of both the stop and the nasal coronals we have accounted for the assimilation to all places of articulation by postulating the absence of a specified Coronal node in those segments. The general restriction of laterals to the Coronal articulator, as argued by Levin (1988) among others, can be expressed in feature geometry by malung Lateral a daughter of the Coronal node. The representation of the coronal lateral, with a specified Coronal node, given in (6) accounts straightfonvardly for its assimilatory possibilities. There is phonological evidence of a significant association between Lateral and Coronal even in the case of palatal and velar laterals. These Dorsal laterals have also a Coronal node, consistent with the claim that the feature lateral is a dependent of Coronal. Independent support for this position can be found in the alternation of the stop and fricative allophones of the voiced obstruents in Catalan, Basque and Spanish. We follow the standard analysis for the general process. Basically, we assurne Continuant Spreading from left to right, i.e. from a segment in a Rime to a segment in the Onset of the following syllable, and [- cont] as the unmarked value. Catalan has both palatal and velar laterals and in this language Continuant Spreading regularly takes place from left to right as can be seen in (7a). The stop appears after a [- cont] segment and after a pause. Instead the fricative appears when the voiced obstruent follows a [+cant] segment as in the last two examples. It has been observed that the [+continuant] value does not spread in homorganic sequences. In the exarnples of (7b) laterals assimilate to the following dental, the homorganicity requirement is met and the spreading of the continuity value is blocked. The same situation holds in the case of the palatal lateral. Clearly, if palatal and velar laterals assimilate to a dental they have to have a Coronal node. (7) a. ambbja blát n3 13a bizl3a b. alkáída belCdía 'envy' 'wheat' 'new' (fem.) 'bishop' 'mayor' 'nice day' We have accounted for the pattern of laterals with respect to assimilatory processes by specifying the Coronal articulator which is predicted to be absent in underlying representations. The consideration of Lateral as a dependent of Coronal justifies the presence of this unmarked articulator. We assume that both palatal and velar laterals are complex segments which branch at the Place node. We will return later to the representation of palatals and the velar lateral as complex segments in relation to gliding processes in Section 3. Let's describe now the assimilatory possibilities of the coronal fricative. As can be seen in (8a) [s] does not assimilate to a labial or velar segment. In some dialects a progressive assimilatory effect can be observed when the coronal fricative follows a sonorant palatal in (8b). The coronal fricative also assimilates to [r] in the example in (&) and the resulting segment has been characterised as a fricative untrilled in Mascar6 (1976). (8) a. mé [S] pa 'more bread' mé [s] camps 'more fields' mt [s fJlors 'more flowers' b. S as 'years' 'grains of garlic' (8) c. mé [z] [rloba +me [Jrloba 'more clothes' The behaviour of the coronal fricative which assimilates only within its primary content node and to palatals can be derived straightforward from the presence of Coronal in the representation of [s] in (9). Strident is relevant only for coronals and we express thls by malung the feature Strident a dependent of the Coronal node. As in the case of [l] in relation to Lateral, the specification on the unmarked articulator in the representation of [s] is justified by malung Strident a dependent of Coronal. If Lateral or Strident are specified for some segment, then the Node Generation convention given in (10) will create automatically a Coronal node. Assuming this we can maintain the unmarked character of Coronal. (16) b. més tr És 'more' 'three' mestr Éns tr sspáls 'more trains' 'three poles' After delinlung of Strident there is no justification for the presence of Corona1 and it deletes because, as the unmarked articulator, it must be absent from underlying representations. The Secondary Articulator, i.e. Dorsal, becomes Pnmary and Strident cannot reassociate. If the affected strident is not a complex segment, the delinked feature can reassociate and this results in no change in the representations. The gliding process affecting a palatal strident is illustrated in (17). Manner Place \ [cont] A parallel process of delateralisation in Occitan has been described by Sauzet (1982) as labialisation. As can be seen in (18) a lateral in syllable final position loses its corona1 and lateral properties when it is followed by another consonant. (18) [val 'it COS~S' [vow f 31 'it will cost' Ea] 'it is necessary' [kow r j] 'it will be necessary' The lateral in Occitan, as in Catalan, is a velar one and we represent them as complex segments with both Coronal and Dorsal articulators. The gliding process consists in an operation which delinks Lateral and the loss of Coronal is a consequence of it. In this way the remaining Dorsal. articulator is interpreted as Pnmary, as represented in (19). I Corond Dorsal 4. Conclusion To summarize, we have accounted for asymmetries in the assimilatory behaviour of coronals by placing Lateral and Strident under Coronal. The observation of dialects which present centdisation of place prior to the assimilation forces us to recognise the double-dependency of these two features. Finally, we examine gliding processes affecting stridents and laterals to show that Configurational Constraints do not suffice and a solution based on Feature Geometry is required. *A first version of this paper was presented at the 1990 GISELLE, Girona. We would like to thank M. Hegarty , J.I. Hualde, J. Mascaró and D. Pulleyblank for their useful comments. This research has been supported by a DGICYT grant PB 89-0324. References Archangeli, D. - D. Pulleyblank (1986) "The Content and Structure of Phonological Representations", ms., University of Arizona & University of Southern California. Avery, P.-K. &ce (1989) "Segment Structure and Coronal Underspecification" , Phonology , 6, 179-200. Cho, Y .Y. (1990) Parameters of Consonantal Assimilution, Doctoral dissertation, Stanford Universi ty. Huaide, J.I. (1988) A Lexical Phonology of Basque, Doctoral dissertation, University of Southern California. Kuen, H. (1932) "El dialecto de Alguer y su posición en la historia de la lengua catalana", Anuari de 1 'Oficina Ronuinica de Lingüística i Literatura, 5, 121-177. Levin (1988) A Place for Lateral in the Feature Geometry , ms., University of Texas. Mascaró, J. (1976) Catalan Phonology and the Phonological Cycle , Doctoral dissertation, MIT. Palmada, B. (1991) La fonologia del cataM i elsprincipis actius. Doctoral dissertation, Universitat Autbnoma de Barcelona. Pairnada, B.- P. Serra (1990) "On the Specification of the Coronal Node: Evidence from Catalan" , paper presented at the 1990 GISELLE, Girona. Paradis, C.-J.F. Prunet (1989) "On coronal transparency", Phonology, 6,3 17-348. Sauzet, P. (1982) "Autour des geminees: des filtres a une critique de segmentalisme", Recherques linguistiques, 10,64123. Yip , M. (1990) "Double -Dependency in Feature Geometry: [murmur] in Shangai and Tibetan and Lateral in Cambodian and Corean", paper presented at the GLOW 1990, Cambridge. UAB-Departamenl de Filologia Catalana Edifici B 081 93 Bellaterra (Barcelona) E-mai1 [email protected]