(PDF) Feature geometry and palatalization

RASPRAVE INSTITUTA ZA HRVATSKI JEZIK I JEZIKOSLOVLJE 32 (2006) UDK 81'344.1 UDK 811.163.42'344.1 UDK 811.162.1'344.1 Izvorni znanstveni članak Rukopis primljen 30. VI. 2006. Prihvaćen za tisak 19. XII. 2006. Małgorzata Ewa Ćavar Sveučilište u Zadru Obala Petra Krešimira IV. 2, HR-23000 Zadar

FEATURE GEOMETRY AND PALATALIZATION The paper examines palatalization data from English, Polish and Croatian in the framework of Feature Geometry. It is shown that palatalization in the discussed data results in a number of phonologically diverse outputs, which cannot be accounted for by using one unique palatalization feature. Instead it is argued that palatalization can be represented by the spreading of either of two independent place specifications of the vowel, or of their combina- tion. This approach implies a threefold typology of palatalization processes. 1. Introduction In this paper I will examine palatalization data from – among others – Eng- lish, Polish, and Croatian. Whereas the assimilatory nature of palatalization has been broadly recognized, it has been a topic of much discussion within the framework of feature geometry since the representation of place of articulation for vowels, unlike for consonants, is not a straightforward issue. In this paper, it will be shown that palatalization may be accounted for if one assumes that it is triggered by two features, which may either spread independently or spread as a unit. This finding requires a reevaluation of the assumptions regarding the representation of front vowels. The paper is organized in the following way. First, in section 2, the two most influential earlier proposals regarding the featural make-up of front vowels will be briefly presented and it will be illustrated how palatalization is repre- sented in these approaches. Section 3 will try to apply these frameworks to a number of palatalization processes in Polish. It will be argued that the palatal- 53 04 Cavar_10.indd 53 12.2.2007 15:12:21 Małgorzata Ewa Ćavar: Feature geometry and palatalization Raspr. Inst. hrvat. jez. jezikosl., knj. 32 (2006), str. 53–74 ization processes in Polish cannot be straightforwardly captured assuming ear- lier proposals. In section 4, an alternative approach will be offered and earlier data will be reanalyzed, and section 5 draws data from Croatian to support the proposal. Section 6 will address some outstanding questions, i.e., affrication of palatalized consonants, and the relative rarity of alveopalatal sounds as an out- put of palatalization. 2. Previous theories of feature geometry with respect to palatali- zation Feature geometry (Clements 1985, Sagey 1986/1990, Hume 1992/1994, Halle 1995, Clements and Hume 1995) has originated from autosegmental phonology (Goldsmith 1990), where the independence of some features with- in a segment is represented by locating the independently acting feature on a separate tier. Feature geometry assumes additionally that if two or more fea- tures act together in phonology, this is represented in theory by grouping these features under a common organizational node. For example, virtually all ap- proaches agree that different place features ([labial], [coronal], [dorsal]) form a larger organizational unit, the Place node, since many languages have pro- cesses of place assimilation where segments assimilate irrespective of wheth- er the source is labial, coronal, or velar, whereas no other features of a given segment are assimilated. Thus, the rule of place assimilation refers directly to the Place node, and the spreading of Place implies the spreading of any feature that is arrayed under this node. (1) Place assimilation in feature geometry C C Place Labial (or Coronal or Dorsal) In the following sections, I will focus on the vowel-consonant assimilation processes, and review the necessary assumptions concerning the representa- tion of the vowels involved in the interaction with consonants from the point of view of two competing models. 54 04 Cavar_10.indd 54 12.2.2007 15:12:22 Małgorzata Ewa Ćavar: Feature geometry and palatalization Raspr. Inst. hrvat. jez. jezikosl., knj. 32 (2006), str. 53–74 1.1. Articulatory model Sagey (1986, 1990) and Halle (1995) assume that all vowels have in their representation the Dorsal node because vowels are articulated with the active involvement of the back part of the tongue. In front vowels, the back part of the tongue is supposed to initiate the forward movement (feature [– back]), whereas there is no such movement for back vowels (feature [+back]). Thus, palatalization is represented in Sagey (1986, 1990) as the spreading of feature [– back] from the vowel onto the consonant (with an automatic insertion of Dorsal node in the consonant). (2) Palatalization of a coronal consonant in Sagey’s (1986/1990) approach C V Place Place (Dorsal) Dorsal Coronal [-back] A consonant with an additional Dorsal node (and attached [-back] feature) represents a consonant with a secondary palatalization. So, this model can ac- count for processes such as, for instance, in Russian, where the result of pala- talization is a consonant with a secondary palatalization. (3) Palatalization in Russian with secondary palatalized consonants as output (a) vdo[v]+a vdo[vj]+e ‘widow’ j (b) rI[b]+a rI[b ]+e ‘fish’ j (c) flo[t]+a flo[t ]+e ‘fleet’ However, it is very often that, apart from the additional secondary articula- tion, palatalization changes the main place of articulation of the consonant to coronal producing anything from dental, over alveolar, post-alveolar, and pre- palatal to palatal sounds. For example, palatalization in English produces pala- 55 04 Cavar_10.indd 55 12.2.2007 15:12:22 Małgorzata Ewa Ćavar: Feature geometry and palatalization Raspr. Inst. hrvat. jez. jezikosl., knj. 32 (2006), str. 53–74 toalveolars [ʃ, ʒ, ʧ, ʤ], both in a lexicalized process as well as a result of a fast speech rule, as in examples below: (4) Palatalization in English (a) [s] — [ʃ] impress — impre[ʃ]n (b) [z] — [ʒ] diffuse — diffu[ʒ]n (c) [t] — [tʃ] Christ — chris[tʃ]n (d) [t] — [ʃ] opt — op[ʃ]n protect — protec[ʃ]n (e) [d] — [ʒ] invade — inva[ʒ]n collide — colli[ʒ]n (f) [d] — [dʒ] grade — gra[ʤ]ual proceed— proce[ʤ]re (g) [s] — [ʒ] Paris — Pari[ʒ]n Caucasus— Cauca[ʒ]n (h) [s] – [ʃ] I miss it - I mi[ʃ] you (i) [z] – [ʒ] I supervi[z]e it — I supervi[ʒ]e your project (j) [t] – [ʧ] I hate it — I ha[ʧ]e you (k) [d] – [ʤ] I nee[d] it — I nee[ʤ] you If we assume that the source of the assimilation, the vowel, is Dorsal (and not Coronal), the change of the target to Coronal is arbitrary. Contrary to intu- ition, palatalization with the change to coronal place of major articulation can- not be seen as an assimilatory process under this approach. An option would be to postulate that the change to the coronal place of articulation is a two-stage process (see 5), first changing the sounds to secondary palatalized ones (5a), than changing secondary palatalized sounds to Coronal (5b). (5) Palatalization with the major change of the articulation place (a) Stage 1: C --> C Dorsal[-back] / V Dorsal[-back] (b) Stage 2: C Dorsal [-back] --> Coronal The notation above states that secondary palatalized consonants become ad- ditionally coronal. Whereas one can design a notation to express this kind of a change, the change is still arbitrary and there is no theory-internal motivation behind it, i.e. the theory neither predicts nor explains it. Also, in English, there is no trace of the intermediate stage. Yet, instead of a simple straightforward no- tation, we are forced in this approach to postulate otherwise unmotivated stages 56 04 Cavar_10.indd 56 12.2.2007 15:12:22 Małgorzata Ewa Ćavar: Feature geometry and palatalization Raspr. Inst. hrvat. jez. jezikosl., knj. 32 (2006), str. 53–74 in the derivation. The criticism of this approach will also be that by definition it is incompatible with the main stream Optimality Theoretic approaches (Prince and Smolensky 1993, McCarthy and Prince 1986, McCarthy 2002), where we cannot refer to intermediate derivational stages and the final result of a phono- logical computation must be delivered in one step. In the following section I will review an alternative theory which treats front vowels as coronal. 1.2. Constriction model Clements (1985) pointed out the inadequacy of the assumption that all vowels are dorsal with respect to palatalization processes. If front vowels trigger the change to the coronal place of articulation in consonants, it would be logical to assume that front vowels contain a feature or node Coronal. This argumentation has been adopted by Hume (1992), Clements and Hume (1995), and many others. In this approach, the nodes are defined as identi- fying not the active articulator but the place of maximal constriction. Back vowels are Dorsal and front vowels have a Coronal node in their representa- tion because the most radical constriction during the production of the vowel is located over the coronal area. Additionaly, Hume (1994) assumes that front vowels are inherently [– anterior]1. This approach is problematic for the account of vowel harmony process- es in many languages. Assuming that the Coronal specification of consonants and the Coronal specification of vowels is the same would be problematic be- cause most of the time consonants do not intervene in vowel harmony pro- cesses. In order to accommodate for the facts of vowel harmony, Coronal specification in vowels occupies a different tier than in consonants, as in (6) below; cf. Clements and Hume (1995). Coronal node in vowels is represent- ed under the V-Place node, and the Coronal node of the consonant is arrayed under the C-Place. 1 This assumption might be problematic for the palatalization processes producing an al- veolar fricative or affricate, as in Polish or Japanese. Yet, in Polish the alveolar [+anterior] sound is an output in a very limited lexicalized alternation being an effect of the whole chain of histori- cal changes. In Japanese, on the other hand, the alveolar sound appears in the context of all high vowels but not all front vowels, thus, one might claim that the alternation is triggered by feature [+high] and thus is a different process than those discussed here. 57 04 Cavar_10.indd 57 12.2.2007 15:12:22 Małgorzata Ewa Ćavar: Feature geometry and palatalization Raspr. Inst. hrvat. jez. jezikosl., knj. 32 (2006), str. 53–74 (6) Representation of a consonant and a front vowel in the constriction model (a) an alveolar consonant (b) a front vowel C V C-Pl C-Pl Coronal V-Pl [+anterior] Coronal Consonants may also have a V-Place specification meant to represent the secondary articulation of consonants, as represented schematically in (7). (7) Secondary palatalization in constriction model C C-Pl Coronal V-Place [+anterior] Coronal Palatalization in this approach is viewed as a place assimilation with Coronal spreading from the vowel. A spreading with a resulting secondary palatalization is represented in (8) below2. 2 Hume (1994:121) assumes that, although front vowels are inherently [– anterior], when vocalic Coronal spreads to a [+anterior] consonant, both the consonant and the vowel are realized as [+anterior]. On the other hand, the presence of [– anterior] specification is crucial in the assimilation involving velar consonants, where [– anterior] value of the resulting consonant comes from the vowel. 58 04 Cavar_10.indd 58 12.2.2007 15:12:22 Małgorzata Ewa Ćavar: Feature geometry and palatalization Raspr. Inst. hrvat. jez. jezikosl., knj. 32 (2006), str. 53–74 (8) Secondary palatalization in the constriction-based approach C V C-Place C-Place Coronal V-Place [+anterior] Coronal The palatalization process resulting in a segment with a coronal major place of articulation is clearly an assimilatory process. However, in order to account for this type of palatalization, it would be necessary to allow an inter-planar spreading, which is not possible for any other features. Here, we have to spec- ify when inter-planar spreading is allowed and when not, and when it actually does occur and when it does not occur although it is not banned. (9) Palatalization with the change of the primary place of articulation, e.g., English C V C-Place C-Place Coronal V-Place [+anterior] Coronal Change of constriction status: yes There are some formal objections to (9). Accepting a spreading between planes just for this one particular feature weakens the theory. A feature geom- etry without inter-planar spreading is a more restrictive and formally more elegant theory. Also, the parameter regarding the change of the constriction status, which in Hume’s model must be included in the statement of the rule in order to account for the spreading either to the Place linked under the Con- sonantal node or to the Place linked under the Vocalic node, is otherwise un- motivated. 59 04 Cavar_10.indd 59 12.2.2007 15:12:22 Małgorzata Ewa Ćavar: Feature geometry and palatalization Raspr. Inst. hrvat. jez. jezikosl., knj. 32 (2006), str. 53–74 3. Palatalization in Polish Hume’s analysis seems to account for the English palatalization data at this point as far as the change of the place of articulation is concerned. How- ever, the theory should be such as to be able to express all possible palataliza- tion processes and exclude all unattested possibilities. In what follows, the data from Polish is discussed and the conclusion will be that we need to propose an- other geometry of the palatalization features than the one assumed in Sagey (1986, 1990) and Halle (1995), or Hume (1992, 1994) and Clements and Hume (1995). In Polish, we can talk about three different regular results of palatalization processes. 2nd Velar Palatalization (cf. Rubach 1984) takes underlying velars /k, g, x/ to surface post-alveolar retroflex-like3 [tš, dž/ž, š], as in (10) below: (10) 2nd Velar Palatalization (a) [k] – [tš] kro[k] ‘step’ kro[tš]yć ‘to step’ (b) [g] – [dž] móz[g] ‘brain’ wymóž[dž]yć ‘to make stupid’ (c) [g] – [ž] wa[g]a ‘scales’ wa[ž]yć ‘to weigh’ (d) [x] – [š] ru[x] ‘movement’ ru[š]yć ‘to move’ One is tempted to assume that Velar Palatalization can be analyzed in the same way as the English data with the change of major place of articulation, as in (9) in the previous section. However, in a moment it will be shown that in Polish this approach is problematic. Another regular palatalization process is Velar Fronting; cf. e.g. Rubach (1984). Velar stops /k, g/ are fronted to pre-velar [kj, gj] (and in a less regular fashion: [x] to [xj]), as illustrated in (11): (11) Velar Palatalization (a) ro[k] ‘year’ Nom.Sg. ro[kj]+em Inst.Sg. (b) blog ‘blog’ Nom.Sg. blo[gj]+em Inst.Sg. 3 Although some researchers count the post-alveolar sounds of Polish among retroflexes (e.g. Hamann 2003), it is important to remember that they are articulatorily quite different from the typical retroflexes like in Dravidian languages. Polish post-alveolars are articulated without the characteristic raising and bunching back of the tip of the tongue. For this reason, they can be also phonetically secondarily palatalized, which often assumed to be impossible for ‘real’ retro- flexes. 60 04 Cavar_10.indd 60 12.2.2007 15:12:22 Małgorzata Ewa Ćavar: Feature geometry and palatalization Raspr. Inst. hrvat. jez. jezikosl., knj. 32 (2006), str. 53–74 (c) rak ‘crab’ Nom.Sg. ra[kj]+i Nom.Pl. (d) dłu[g]+a ‘long’ Nom.Sg.Fem. dłu[gj]+i Nom.Sg.Masc. (e) ma[x]+ać ‘to wave’ wyma[xj]+iwać ‘to wave’ Iterative Again, without looking at the whole system of palatalization processes in Polish, Velar Palatalization might potentially be accounted for as a spreading of the Coronal specification arrayed under the V-Place; cf. (8). Finally, Coronal Palatalization (cf. Rubach 1984) needs to be mentioned, where dental stops and fricatives [t, d, s, z] (but not affricates), as well as the dental nasal [n], alternate with prepalatals [ʨ,ʥ, ɕ, ʑ, ɲ]4. (12) Coronal Palatalization (a) masa ‘mass’ ma[ɕ]+e Dat./Loc.Sg. (b) brąz ‘brown color’ brą[ʑ]+e Loc./Voc.Sg. (c) brat ‘brother’ bra[ʨ]+e Loc./Voc.Sg. (d) moda ‘fashion’ mo[ʥ]+e Dat./Loc.Sg. (e) Jan proper name Ja[ɲ]+e Loc./Voc.Sg. At first sight, one could be tempted to analyze Coronal Palatalization as the spreading of Coronal to the C-Place of the consonant. However, then the prediction would be that the output of Coronal Palatalization should not be different from the output of 1st Velar Palatalization. In fact, both the post-al- veolar results of 1st Velar ([š, ž, tš, dž]), and the prepalatal effects of Coronal Palatalization ([ʨ,ʥ, ɕ, ʑ, ɲ]) are coronal, non-anterior. Yet they differ in that prepalatals are ‘inherently palatalized’ (cf. Keating 1991, Halle and Ste- vens 1989), that is, the mid part of the tongue behind the blade is raised to- wards the hard palate, the configuration that is particular for every second- ary palatalized segment. Post-alveolars are also coronal, non-anterior, so the major place of articulation is the same, but in their articulation the tongue lies flat in the mouth cavity. 4 The alternation of liquids, i.e., [r] changing to [ž] in the context of a front vowel, as well as [w] alternating with [l] is often treated with the change resulting in prepalatals. I do not dis- cuss these alternations here because they need an independent account under any framework; cf. Ćavar (2004). 61 04 Cavar_10.indd 61 12.2.2007 15:12:22 Małgorzata Ewa Ćavar: Feature geometry and palatalization Raspr. Inst. hrvat. jez. jezikosl., knj. 32 (2006), str. 53–74 (13) Post-alveolars and prepalatals (after Wierzchowska 1967, 1980) (a) post-alveolar [š, ž, tš, dž] (b) prepalatal [ʨ,ʥ, ɕ, ʑ, ɲ] In terms of the constriction-based model, prepalatals have two Coronal specifications: one under the C-node, and another under the V-node. As a con- sequence, adopting the representation of the front vowel as in Clements and Hume (1995) would require that the Polish Coronal Palatalization is regard- ed as spreading of the Coronal node at the same time to the C-Place and the V- Place of the consonant, that is, it would require two independent operations at the same time. (14) Coronal palatalization as a double spreading. C V C-Place C-Place Coronal (V-Place) V-Place [+anterior] Coronal [-anterior] Thus, we end up again postulating a two-step solution. The data from Pol- ish indicates that in order to distinguish between the different types of process- es in Polish we actually need two ‘palatalization features’ to spread from the vowel, and a common node to dominate them for the cases when the two fea- tures act together. 62 04 Cavar_10.indd 62 12.2.2007 15:12:22 Małgorzata Ewa Ćavar: Feature geometry and palatalization Raspr. Inst. hrvat. jez. jezikosl., knj. 32 (2006), str. 53–74 4. Proposal In order to express this idea that prepalatals originate from the spreading of two ‘palatalization’ features, we would have to modify either the Articulator model or the Constriction model with regard to the featural make-up of front vowels. Within the Articulator model, we would need to assume that front vowels are both Dorsal [-back] and Coronal [-anterior]. A similar assumption has been adopted in Rubach (1993). In Rubach’s analysis of Slovak it is assumed that front vowels are dorsal and at a fairly early stage of derivation, before the application of palatalization rules, they are redundantly specified as [-anterior] at the Coronal node. Instead of assuming derivational steps, in principle one could assume that front vowels are both coronal[-anterior] and dorsal – by definition; cf. Cavar (1997). Yet this framework might be problematic for the exact same reasons why Clements and Hume (1995) assumed a division between the consonantal and vocalic place. This approach predicts that front vowels will interact with consonants in vowel harmonies. In fact, Hume (1992, 1994) described one such case in Maltese. Yet, it is more often that the major place of articulation of intervening consonants is transparent in vowel harmony processes, thus, there must be a theoretical tool to account for the distinction between consonantal and vocalic features. The option would be to adopt the constriction model with all the benefits of the distinction between vocalic and consonantal place features, and modify it in the way that front vowels are additionally specified as Coronal at C-Place (in- stead of having an empty C-Place). (15) Front vowel in the modified model V C-Place Coronal V-Place [-anterior] Coronal [-anterior] 63 04 Cavar_10.indd 63 12.2.2007 15:12:22 Małgorzata Ewa Ćavar: Feature geometry and palatalization Raspr. Inst. hrvat. jez. jezikosl., knj. 32 (2006), str. 53–74 Within this approach we would analyze Coronal Palatalization as the spread- ing of the whole Place specification (including C-Place and V-Place) from the front vowel, as illustrated in (16): (16) Coronal Palatalization C V C-Place C-Place Coronal Coronal V-Place [+anterior] [-anterior] Coronal [-anterior] Further, 1st Velar Palatalization in Polish would be seen as a spreading of the Coronal node directly dominated by C-Place from the vowel alone. (17) 1st Velar Palatalization C V C-Place C-Place = Dorsal Coronal V-Place [-anterior] Coronal [-anterior] This proposal leaves room for the account of the secondary palatalization without the change of the major place of articulation, and explains processes such as the palatalization in Russian or Velar Fronting in Polish. 64 04 Cavar_10.indd 64 12.2.2007 15:12:22 Małgorzata Ewa Ćavar: Feature geometry and palatalization Raspr. Inst. hrvat. jez. jezikosl., knj. 32 (2006), str. 53–74 (18) Velar Fronting C V C-Place C-Place = Dorsal V-Place Coronal Coronal [-anterior] [-anterior] Finally, we would have to revise our approach to the palatalization in Eng- lish. English palatoalveolars are articulated with the inherent raising of the tongue towards the hard palate, similarly like in prepalatals (though the dis- placement of the tongue from the rest position is less extreme than in prepal- atals). Thus, more true would be to claim that the English palatoalveolars are represented with two coronal specifications, one for the place of articulation, and the other for the raising of the coronal part of the tongue towards the hard palate. (19) English palatoalveolars C C-Pl Coronal V-Place [-anterior] Coronal [-anterior] The palatalization in English should be seen from this perspective as a spreading of the whole Place node of the vowel, because otherwise we should obtain sounds which are non-anterior but articulated without the ‘inherent pala- talization’, i.e., comparable to Polish series [š, ž, tš, dž]: 65 04 Cavar_10.indd 65 12.2.2007 15:12:23 Małgorzata Ewa Ćavar: Feature geometry and palatalization Raspr. Inst. hrvat. jez. jezikosl., knj. 32 (2006), str. 53–74 (20) English palatalization C V C-Place C-Place Coronal Coronal V-Place [+anterior] [-anterior] Coronal [-anterior] 5. Palatalization in Croatian One possible criticism of this approach is that Polish palatalization process- es belong partly in the realm of morphophonology, that is, their application de- pends on the choice of the morphemes. The next step would be to claim that all the processes are actually lexicalized, and particular for Polish only, and there is no need to propose a theory to explain their potential output. I will now turn to the data from Croatian and show that the Croatian data also requires the as- sumption of two ‘palatalization’ features. Although Polish and Croatian are ge- netically related, the genesis of prepalatals in the two languages is independent (Dalewska-Greń 2002: 96ff). Thus, if two languages show independently the same behavior, we want to claim that the phonological theory needs to explain the emergence of the regular effect. Croatian, like Polish, has both non-secondarily-palatalized post-alveolars [š, ž, tš, dž] and prepalatals [ʨ, ʥ] (see Miletić 1933 for the relevant x-ray trac- ings which help to verify the classification of sounds in terms of place of ar- ticulation). In the Croatian literature (e.g. Težak and Babić 1996), we will find sometimes classifications of [š, ž, tš, dž] as post-alveolar or prepalatal sounds and [ʨ, ʥ] are described as real palatals. Barić et al. (1990:25) describe both [š, ž, tš, dž] and [ʨ, ʥ] as palatal sounds, the former as apical, the latter as lam- inal. This use of labels only partly corresponds to the broadly accepted Eng- lish terminology that is adopted in this article. Thus, we refer to [š, ž, tš, dž] as to post-alveolars (without specifying further), and to [ʨ, ʥ] as prepalatals (cf. e.g. Brozović 1991), since it seems that these terms in the best way correspond to the definitions of terms as they are used in the phonological English-speak- 66 04 Cavar_10.indd 66 12.2.2007 15:12:23 Małgorzata Ewa Ćavar: Feature geometry and palatalization Raspr. Inst. hrvat. jez. jezikosl., knj. 32 (2006), str. 53–74 ing literature. For the overview of the relevant terminology from the phonet- ic perspective, see Keating (1991) and, from the phonological perspective Hall (1997). Prepalatals are a result of historical palatalization process, and post-al- veoalars may be an output of a synchronic alternation, cf. also diachronically Mihaljević (1991). The Croatian synchronic alternation is parallel to Polish 1st Velar Palatalization; cf. (21). (21) Croatian Velar Palatalization juna[k] ‘hero’ Nom.Sg. juna[tš]e Voc.Sg. knji[g]a ‘book’ Nom.Sg. knji[ž]ewni ‘literary’ [du[x] ‘ghost’ Nom.Sg. du[š]e Voc.Sg. In Croatian, there is no parallel of Polish Coronal Palatalization, yet there are prepalatals [ʨ, ʥ], and the two groups of sounds clearly have different fea- tural representation. Given their phonetic similarity to Polish sounds, one can safely assume that whereas [tš, dž, š, ž] are coronal [-anterior] under the C- Place, [ʨ, ʥ] additionally have a V-Place with a Coronal specification. Croatian prepalatals historically originated from the sequences of proto- Slavic *tj, *dj, the change going back to Old Slavic. This is different from Pol- ish, because the same original sequences are rendered in Modern Polish as [ts, dz]; see (22a). On the other hand, Polish prepalatals clearly have no Croatian prepalatal counterparts, as exemplified in (22b). (22) Croatian Polish gloss (a) no[ʨ] no[ts] ‘night’ Nom.Sg. svije[ʨ]e świe[ts]e ‘candles’ Nom.Pl. pre[ʥ]a przę[dz]a ‘yarn’ Nom.Sg. izme[ʥ]u mię[dz]y ‘between’ (b) [t]ijesto [ʨ]asto ‘dough’ Nom.Sg. [dj]eca [ʥ]eci ‘children’ Nom.Pl. ra[d]i ra[ʥ]e ‘advise’ Dat.Sg. Unlike Croatian, Polish prepalatals originated from secondarily palatalized den- tals, and the change was completed by 16th century only (Klemensiewicz 1985). 67 04 Cavar_10.indd 67 12.2.2007 15:12:23 Małgorzata Ewa Ćavar: Feature geometry and palatalization Raspr. Inst. hrvat. jez. jezikosl., knj. 32 (2006), str. 53–74 The Croatian palatalization as in (21) should be viewed as a spreading of the Coronal node underneath the C-Place without the V-Place. If it were a spread- ing involving the V-Place alone, the result would be a secondary palatalized ve- lar, and if the whole specification of the vowel spread, the outcome would be a prepalatal, which is not the case. Croatian has to differentiate between the outputs of Velar Palatalization and the prepalatals, thus, a more complex representation of a front vowel than in Hume (1994), Sagey (1990), or Clements and Hume (1995) seems necessary. There are actually more languages which differentiate between two types of non-anterior coronals where one is articulated with the raising of the blade of the tongue towards the hard palate, and the other non-anterior sound does not have this ‘inherent palatalization’, and where at least one member of the pair is derived in the context of a front vowel. One more such example is Swedish. A prepalatal voiceless fricative [ɕ], a coronal [-anterior] sound, appears always in the context of a front vowel and contrasts with a retroflex sound, which is also coronal [-anterior]. Clearly, in order to get a prepalatal as an effect of palatal- ization, we need to explicitly specify that the outcome has additionally – apart from coronal [-anterior] specification under the C-Place node – also a coronal specification under the V-Place. Otherwise, it would be unclear why the output is a prepalatal and not a retroflex. Summing up, Polish is not the only language where we would need to as- sume a representation of a front vowel with a double coronal specification, thus, the data problematic for the earlier theories cannot be dismissed as a idio- syncratic Polish phenomenon. 6. Residual problems The spreading of any specification under the place node does not account for the fact that, whereas fricatives undergoing palatalization do not change the manner of articulation, stops alternate with affricates. In actual fact, feature ge- ometry does not have theoretical tools to explain affrication. The problem might find a formal solution if we assumed, following Lom- bardi (1990) and Sagey (1986), that affricates contain a double specification for the feature [continuant]: they are both [−continuant] and [+continuant]. Affri- cation in this case might be considered to be a spreading of [+continuant] from the vowel, as illustrated in (23). 68 04 Cavar_10.indd 68 12.2.2007 15:12:23 Małgorzata Ewa Ćavar: Feature geometry and palatalization Raspr. Inst. hrvat. jez. jezikosl., knj. 32 (2006), str. 53–74 (23) Affrication C V [-continuant] [+continuant] However, the structural description of the rule is also met when the con- dition of the rule is a back vowel, yet the spreading does not occur from back vowels. Low or back vowels do not seem to have the tendency to affricate adja- cent stops, and it is not clear why a spreading of continuancy should be limited only to the situations when the trigger is a front vowel. One cannot say, for in- stance, that front vowels are more continuant than back vowels. Another approach was suggested by Padgett (Padgett 1995), who assumes that feature [continuant] is arrayed under place features. Thus, a spreading re- ferring exclusively to Coronal would be impossible, and the prediction would actually be that if the place of articulation in the consonant changes to coronal, the manner of articulation must also change to an affricate or a fricative. This would motivate a concomitant mutation of Coronal and manner feature in pala- talized consonants. The obvious problem in this approach would be of course the secondary palatalization, where manner of articulation does not change, i.e., the output of palatalization remains a stop. Bhat (1978) also reports palataliza- tion with the change of the major place of articulation to a prepalatal stop (Bhat 1978, after Miller and Davis 1963). And last but not least, recent studies (e.g., Kim 2001) present arguments against feature [+continuant] in the representa- tion of affricates altogether5. The two legitimate operations in feature geometry are a spreading of a fea- ture (or a node) and a delinking of a feature. When talking about affrication in the context of a front vowel, we cannot regard it as a spreading from a front vow- el because there is no ‘affrication’ or ‘stridency’ feature in the vowel, but rather it would have to be a feature insertion. There is in principle nothing wrong about feature insertion as long as it is phonetically motivated. Yet the question is then, providing affrication is a phonological effect, how phonology knows which fea- ture insertion is phonetically motivated and which is not, consequently, how it can be distinguished between a phonetically legitimate insertion and the fea- ture insertion which is not phonetically motivated. The problem of distinguish- ing between possible and impossible sound changes is supposed to be the core 5 Earlier researchers assumed that affricates should be regarded as strident stops; cf., e.g., Rubach (1994). 69 04 Cavar_10.indd 69 12.2.2007 15:12:23 Małgorzata Ewa Ćavar: Feature geometry and palatalization Raspr. Inst. hrvat. jez. jezikosl., knj. 32 (2006), str. 53–74 of phonological theory, and should be formalized. Yet, the theory of feature ge- ometry does not contain a sub-theory of possible feature insertion. It is unquestionable that affrication of a consonant in the context of a front vowel “makes sense” phonetically. Lahiri and Evers (1991), cf. also Kim (2001), motivated the appearance of affricates instead of stops as a side-effect of the characteristic articulation of the palatalized consonant with an off-glide [j] release: “Coronal consonants have relatively more energy in the higher frequencies than in the lower frequencies (Lahiri et al. 1984: 402). If, in addition, there is an off-glide [j] release for the [+high] palatalized coronal consonants, then there will be a greater increase in the higher frequencies, causing a concentration of energy in the high frequency range – a characteristic of strident segments.” (La- hiri and Evers 1991:95) This passage aptly explains the rise of stridency in phonetic terms. Howev- er, it is unclear how this account should be formalized within Feature Geome- try, and how acoustic properties of the sound should influence the structure of a sound without prior assuming the existence of perceptual features. Another explanation, this time in terms of articulation, was hinted in Lahiri and Evers (1991). It has been proposed that stridency emerges because of the change to the palato-alveolar region, where the unmarked articulation of all obstruents is with stridency; cf. Lahiri and Blumstein (1984: 142). It is true that in languages of the world non-anterior obstruents tend to be affricates or fricatives. Yet, there are languages which have post-alveolar stops in their inventory (cf. Ladefoged and Maddieson 1996), and Bhat (1978) quotes one example of a language where palatalization produces a prepalatal stop, namely, Acoma (Bhat 1978 after Mill- er and Davis 1963). Thus, feature geometry cannot explain affrication. Another problem is why it is so seldom that palatalization results in post-al- veolar segments, relatively more seldom than the palatalization to a palatoalve- olar or secondary palatalization. The answer to this question has to necessarily refer to the theory of inventories and the theory of markedness, feature geome- try leaves this question unanswered. 7. Conclusions The analysis in the approach of Sagey, as it was proposed originally in 1986 (Sagey 1986, 1990), cannot adequately account for the data whenever we have a palatalization with the change of the major articulation place to coronal. The model would be able to express the assimilatory nature of the process only if 70 04 Cavar_10.indd 70 12.2.2007 15:12:23 Małgorzata Ewa Ćavar: Feature geometry and palatalization Raspr. Inst. hrvat. jez. jezikosl., knj. 32 (2006), str. 53–74 we adopted the amendments by Rubach (1993) or Ćavar (1997). The former ap- proach cannot be adopted within an OT approach, the latter models would be problematic for the account of vowel harmony processes. In this article a modification of Hume-Clements model has been proposed as to include a Coronal specification on the C-Place tier in the representa- tion of the vowel. This model predicts three possible kinds of a spreading: the spreading of the Coronal node from the C-Place, the spreading of the Coronal from the V-Place node, and the spreading of the whole Place node. In the orig- inal Clements and Hume’s approach, the front vowel may spread in one op- eration either to the vocalic or to the consonantal Place. The Clements-Hume model predicts that the changes in a natural language changing both the main articulation to coronal [-anterior] and adding the secondary palatalization are a pure coincidence, and as such should not reoccur cross-linguistically. This prediction is not borne out, as it has been demonstrated in earlier sections. This paper does not propose a solution to other issues problematic for fea- ture geometry, such as the affrication concomitant with the change of the place of articulation in palatalization, or the relative rarity of post-alveolars as an ef- fect of palatalization. Instead, it is suggested that these problems need to be ex- plained with reference to perception. References: BARIĆ, EUGENIJA; MIJO LONČARIĆ; DRAGICA MALIĆ; SLAVKO PAVEŠIĆ; MIRKO PETI; VESNA ZEČEVIĆ; MARIJA ZNIKA 1990. Gramatika hrvatskoga književnog jezi- ka. Zagreb: Školska knjiga. BHAT, D.N.S. 1978. A General Study of Palatalization. In: Greenberg, Joseph H. (ed.), Universals of Human Language, Vol. 2. Stanford, Ca.: Stanford University Press. 47–91. BROZOVIĆ, DALIBOR 1991. Fonologija hrvatskoga književnog jezika. In: Stjepan Babić, Dalibor Brozović, Milan Moguš, Slavko Pavešić, Ivo Škarić, Stjep- ko Težak, Povijesni pregled, glasovi i oblici hrvatskoga književnog jezika. Zagreb: Globus. 379–452. CLEMENTS, GEORGE N. 1985. The Geometry of Phonological Features. Phonolo- gy Yearbook, 2, 225–252. CLEMENTS, GEORGE N.; ELIZABETH V. HUME 1995. The Internal Organization of Speech Sounds. In: Goldsmith, John A. (ed), The Handbook of Phonologi- cal Theory. Oxford: Blackwell. 24–306. ĆAVAR, MAŁGORZATA 1997. Coronal Sounds and Palatalization: An Analysis of Polish and English. Warsaw University M.A. thesis. 71 04 Cavar_10.indd 71 12.2.2007 15:12:23 Małgorzata Ewa Ćavar: Feature geometry and palatalization Raspr. Inst. hrvat. jez. jezikosl., knj. 32 (2006), str. 53–74 ĆAVAR, MAŁGORZATA 2004. Palatalization in Polish. An Interaction of Articula- tory and Perceptual Factors. University of Potsdam Ph.D. diss. DALEWSKA-GREŃ, HANNA 2002. Języki słowiańskie. Warszawa: Wydawnictwo Naukowe PWN. GOLDSMITH, JOHN A. 1990. Autosegmental and Metrical Phonology. Oxford, UK, Cambridge, Ma.: Blackwell. HALL, T. ALAN 1997. The Phonology of Coronals. Current Issues in Linguistic Theory, vol. 149. Amsterdam: John Benjamins. HALLE, MORRIS 1995. Feature Geometry and Feature Spreading. Lingustic Inqu- iry, 26 (1), 1–46. HALLE, MORRIS; KENNETH STEVENS 1989. The Postalveolar Fricatives of Polish, Ms. MIT. HAMANN, SILKE 2003. The Phonetics and Phonology of Retroflexes. The Hague: Holland Academic Graphics. HUME, ELIZABETH V. 1992. Front Vowels, Coronal Consonants, and their Inte- raction in Nonlinear Phonology. Cornell University PhD. diss. HUME, ELIZABETH V. 1994. Front Vowels, Coronal Consonants, and their Inte- raction in Nonlinear Phonology. New York: Garland. KEATING, PATRICIA 1991. Coronal Places of Articulation. In: Carole Paradis and Jean-Françoise Prunet (eds), The Special Status of Coronals. Internal and External Evidence. San Diego, Ca.: Academic Presss. 29 – 48. KIM, HYUNSOON 2001. A Phonetically Based Account of Phonological Assibila- tion. Phonology, 18, 81–108. KLEMENSIEWICZ, ZENON 1985. Historia języka polskiego. Warszawa: Państwowe Wydawnictwa Naukowe. LADEFOGED, PETER; IAN MADDIESON 1996. The Sounds of the World’s Languages. Cambridge, Ma.: Blackwell. LAHIRI, ADITI; SHEILA E. BLUMSTEIN 1984. A Re-evaluation of the Feature Coro- nal. Journal of Phonetics, 12, 133–146. LAHIRI, ADITI; LETITIA GEWIRTH; SHEILA E. BLUMSTEIN 1984. A Reconsideration of Acoustic Invariance for Place of Articulation in Diffuse Stop Consonants. Evidence from a Cross-language Study. Journal of the Acoustic Society of America, 76, 391–404. LAHIRI ADITI; VICTOR EVERS 1991. Palatalization and Coronality. In: Carole Pa- radis and Jean-François Prunet (eds), The Special Status of Coronals. Inter- nal and External Evidence. San Diego: Ca: Academic Press. 79–100. LOMBARDI, LINDA 1990. The Nonlinear Organization of the Affricate. Natural Language and Linguistic Theory, 8, 374–425. 72 04 Cavar_10.indd 72 12.2.2007 15:12:23 Małgorzata Ewa Ćavar: Feature geometry and palatalization Raspr. Inst. hrvat. jez. jezikosl., knj. 32 (2006), str. 53–74 MCCARTHY, JOHN 2002. A Thematic Guide to OT. Cambridge: Cambridge Uni- versity Press. MCCARTHY, JOHN; ALAN PRINCE 1986. Prosodic Morphology. Tech Rept Rutgers University Center for Cognitive Science, New Brunswick, NJ. Report No RuCCS-TR-32. MIHALJEVIĆ, MILAN 1991. Generativna fonologija hrvatske redakcije crkveno- slavenskog jezika. Zagreb: Filozofski fakultet, Odsjek za opću lingvistiku i orijentalne studije. MILETIĆ, BRANKO 1933. Izgovor srpskohrvatskih glasova: eksperimentalno-fo- netska studija. Beograd: Srpska kraljevska akademija. MILLER, WICK R.; IRVINE DAVIS 1963. Proto-Keresan Phonology. IJAL, 29(4), 310–330. PADGETT, JAYE 1995. Stricture in Feature Geometry. Stanford, Ca.: CSLI Pu- blications. PRINCE, ALAN; PAUL SMOLENSKY 1993. Optimality Theory: Constraint Interacti- on in Generative Grammar. Tech. Rept. Rutgers University center for Co- gnitive science, New Brunswick, NJ. Report No RuCCS-TR-2. RUBACH, JERZY 1984. Cyclic and Lexical Phonology: the Structure of Polish. Dordrecht, Holland, Cinnaminson, N.J.: Foris Publications. RUBACH, JERZY 1993. The Lexical Phonology of Slovak. Oxford: Clarendon Press. RUBACH, JERZY 1994. Affricates as Strident Stops in Polish. Linguistic Inquiry, 25, 119–143. SAGEY, ELIZABETH 1986. The Representation of Features and Relations in Non- linear Phonology. MIT Ph.D. diss. SAGEY, ELIZABETH 1990. The Representation of Features in Non-linear Phono- logy: the Articulator Node Hierarchy. New York: Garland. TEŽAK, STJEPKO; STJEPAN BABIĆ 1996. Gramatika hrvatskoga jezika. Priručnik za osnovno jezično obrazovanje. Zagreb: Školska knjiga. WIERZCHOWSKA, BOŻENA 1967. Opis fonetyczny języka polskiego. Warszawa: Państwowe Wydawnictwo Naukowe. WIERZCHOWSKA, BOŻENA 1980. Fonetyka i fonologia języka polskiego. Wro- cław: Zakład Narodowy im. Ossolińskich. 73 04 Cavar_10.indd 73 12.2.2007 15:12:23 Małgorzata Ewa Ćavar: Feature geometry and palatalization Raspr. Inst. hrvat. jez. jezikosl., knj. 32 (2006), str. 53–74 Geometrija obilježja i palatalizacija Sažetak U članku se istražuje palatalizacija u poljskome, hrvatskome i engleskome jeziku u okvi- ru geometrije obilježja. Raspravlja se o tome kako palatalizaciju možemo pojmiti kao šire- nje obilježja mjesta tvorbe izravno ispod konsonantskoga mjesnog čvora (C-Place node) ili kao širenje mjesta tvorbe ispod vokalnoga mjesnog čvora (V-Place node), odnosno njiho- vu kombinaciju. Rasprava počinje kratkim pregledom dvaju najutjecajnijih pristupa palatalizaciji unu- tar geometrije obilježja: artikulacijskoga modela (Articulatory Model, Sagey 1986 i 1999, Halle 1995) i modela stezanja (Constriction Model, Hume 1992, 1994, Clements i Hume 1995). Potom slijedi prikaz podataka o palatalizaciji u poljskome jeziku. Dokazano je kako nijedan od modela nije u potpunosti primjenjiv pri analizi tih podataka. Artikulacijski mo- del ne može predvidjeti pravilnu izmjenu mjesta artikulacije konsonanata u koronalni glas u okružju prednjih vokala. Model stezanja, s druge strane, predviđa dvije vrste rezultata tije- kom procesa palatalizacije: sekundarno palatalizirane glasove (bez promjene glavnoga mje- sta artikulacije) i promjenu glavnoga mjesta artikulacije koronalnih glasova (bez sekundarne palatalizacije). Model ne predviđa mješoviti učinak, a poljski prednjonepčani glasovi nasta- li palatalizacijom ipak su rezultat kombinacije dviju promjena, tj. promjene glavnoga mjesta artikulacijske promjene i dodavanja sekundarne palatalizacije. Stoga palatalizacija u polj- skome jeziku predstavlja problem i u primjeni modela stezanja. Slični se problemi pojavlju- ju i kod analize palatalizacije u standardnome hrvatskom jeziku, što pokazuje da te promje- ne nisu tek pojedinačne osobitosti poljskoga ili hrvatskoga jezika, već su indikativne s obzi- rom na mehanizme koji utječu na palatalizaciju u pojedinim jezicima te ukazuju na potrebu za sustavnim rješenjem koje će biti primijenjeno na analizu primjera iz obaju jezika. U članku se zagovara potpuno nov pristup. Dokazuje se da prednji vokal ima dvostruku koronalnu (prednju) specifikaciju mjesta tvorbe. Dio bi se palatalizacijskih procesa iz kojih proizlazi sekundarna palatalizacija konsonanata (primjerice, pomicanje jedrenika u poljsko- me) valjao pripisati širenju obilježja vokalnoga mjesnoga čvora dalje od vokala. Dio je pa- latalizacijskih procesa i stvaranja ne-prednjih koronalnih glasova koji se artikuliraju bez po- dizanja jezika prema tvrdome nepcu u poljskome i hrvatskome (primjerice prva velarna pa- latalizacija u poljskome i velarna palatalizacija u hrvatskome), uzrokovan širenjem koronal- noga (prednjeg) mjesta tvorbe ispod konsonantskoga mjesnog čvora pri artikulaciji vokala. Konačno, širenje cijeloga mjesnog čvora zahtijeva promjenu u koronalne i palatoalveolarne konsonante (koronalna palatalizacija u poljskome). Rad na kraju donosi kratak pregled mogućih nerazriješenih pitanja o načinu pred- stavljanja palatalizacijskih procesa unutar pristupa geometrije obilježja. Ključne riječi: fonologija, geometrija obilježja, palatalizacija, poljski jezik, hrvat- ski jezik, engleski jezik Keywords: phonology, feature geometry, palatalization, Polish, Croatian, English 74 04 Cavar_10.indd 74 12.2.2007 15:12:23