Hostname: page-component-6898cfcb55-zkb6p Total loading time: 0 Render date: 2026-11-11T10:29:13.754Z Has data issue: false hasContentIssue false
  • English
  • Français

The Representation of Affricates in Cimbrian German

Published online by Cambridge University Press:  07 February 2012

Tracy Alan Hall
Tracy Alan Hall*
Affiliation:
Indiana University
*
Department of Germanic Studies, Ballantine Hall 644, Indiana University, 1020 Kirkwood Avenue, Bloomington, Indiana 47405–7103, USA, [tahall2@indiana.edu]
Rights & Permissions [Opens in a new window]

Abstract

It is argued that Cimbrian German requires reference to the natural class of affricates and fricatives and that this grouping is captured by analyzing both sets of sounds as [+continuant]. This finding is problematic for theories that treat affricates as strident stops without a [+continuant] component.*

Information

Type
ARTICLES
Information
Journal of Germanic Linguistics , Volume 24 , Issue 1 , 07 February 2012, pp. 1 - 22
Copyright
Copyright © Society for Germanic Linguistics 2012

1. Introduction

The present study focuses on the phonological behavior of affricates in a regional German dialect spoken in Northern Italy (Cimbrian German; Bacher 1905, Tyroller 2003). It is shown that the context for various allomorphs in the nominal and verbal inflexion of Cimbrian German requires the analysis of affricates in that dialect as [+continuant]. This claim is significant because it conflicts with the representation of affricates proposed by linguists such as La Charité (1993), Rubach (1994), Clements (1999), Kim (2001), and Kehrein (2002) (on the basis of Jakobson et al. 1952), which sees affricates as strident stops without a [+continuant] component.

In section 2, I summarize some of the competing models for affricates that have been proposed in the crosslinguistic literature on theoretical phonology. In section 3, I present the relevant data from Cimbrian German, and in section 4, I discuss several open research questions.

2. The Structure of Affricates

In recent years, a number of competing featural models have been proposed for the representation of affricates (see Lin 2011 for a recent survey). In this section, I discuss briefly the Contour Segment Representation, the Complex Segment Representation, and the Strident Stop Representation, focusing on two of the predictions made by the latter model.

According to the first model, affricates are (monosegmental) contour segments in the sense that they consist of a single root node (with the major class features [–son(orant)] and [−cons(onantal)]) that dominates [−cont(inuant)] and [+cont(inuant)] in that linear order. Given this structure, affricates as represented in 1a are distinct from stops in 1b and fricatives in 1c because the latter two segment types only have one specification each for [±cont].

  1. (1)

First proposed by Sagey (1986), the Contour Segment Representation has been argued to capture edge effects: Phonologically, affricates pattern with the stops on their left edge and with fricatives on their right edge. For example, the English regular plural requires an epenthetic vowel between an affricate or a coronal fricative and /z/, as in churches, bridges, bushes. Since there is no epenthesis if the stem ends in a coronal stop (as in cats), Sagey (1986) concludes that epenthesis depends on a representation like the one in 1a, in which affricates are fricatives on their right side.

The Contour Segment Representation in 1a has been challenged by a number of researchers (see, among others, Hualde 1988, Lombardi 1990). According to Lombardi 1990, the [−cont] and [+cont] autosegments (which are sometimes interpreted as the corresponding privative features [stop] and [fricative], respectively) are situated on different autosegmental tiers. In the representation in 2a, affricates are complex segments in the sense that they are both [stop] and [cont]. In 2b,c, stops and fricatives are represented in that model with the features [stop] and [cont], respectively.

  1. (2)

Within this model, the representation in 2a predicts that affricates behave phonologically as stops or fricatives on either side. The argument for 2a and against 1a is said to derive support from the existence of antiedge effects. For example, in Turkish, both stops and affricates (but not fricatives) are devoiced at the right edge of a syllable, i.e., [−cont]→ [−voice]/__]σ. This type of rule is problematic for 1a because this representation incorrectly predicts that syllable-final affricates should not devoice. This prediction follows from the representation of affricates in 1a as [−cont] is not situated at the right edge of a syllable. Lombardi (1990) concludes that the representation in 2a is correct because it predicts that affricates can behave like stops or fricatives on either side.Footnote 1

Both the Contour Segment Representation and the Complex Segment Representation have been rejected by proponents of the Strident Stop Representation (see, among others, La Charité 1993, Rubach 1994, Clements 1999, Kim 2001, Kehrein 2002, and Hall 2004 for Modern Standard German, who base their models on the earlier one proposed by Jakobson et al. 1952). According to the representation in 3a, affricates are [−cont], but there is no [+cont] component. Instead, affricates are distinguished from the corresponding stops with the feature [±stri(dent)]: Affricates are [+stri] and stops are either not marked for the feature [stri] at all, as in 3b, or they are [–stri].

  1. (3)

Strident fricatives such as /s z/ require that the representation in 3c be augmented with the feature [+stri]. There is general—although not unanimous—agreement among the linguists cited above that [+stri] only describes sibilant sounds. Thus, sibilant affricates are segments such as /ts dz /, sibilant fricatives include sounds such as /s z /, and [−stri] are the fricatives /θ f ç x/. The status of nonstrident affricates such as [pf] under 3a is discussed below.

The three models presented above make a number of falsifiable claims, two of which are presented in 4.

  1. (4) Two predictions made by the Strident Stop Representation

    1. a. Strident affricates can contrast with stops at the same place of articulation, but nonstrident affricates cannot.

    2. b. Affricates and fricatives do not constitute a natural class.

Prediction 4a says that it is possible to have a language that contrasts [t] and [ts], for example, in which both sounds are either dental or alveolar. By contrast, nonstrident affricates such as /pf/ (in Modern Standard German; henceforth MSG), /tθ/ (in the Nilotic language Shilluk spoken in Sudan), and /kx/ (in Swiss German) never occur together in the same language with stops at the same place (Kehrein 2002). Put differently, in languages with /t/ and /tθ/ or /p/ and /pf/, there is always a minor place contrast that distinguishes the two sounds. For example, in MSG /p/ is bilabial and /pf/ is labio-dental; in Shilluk, /t/ is lamino-alveolar and /tθ/ is apico-dental (Kehrein 2002:27). This means that in a language such as Shilluk, /t/ and /tθ/ are both [–son, –cont, –stri], and an independent feature (for instance, [±distr(ibuted)]) distinguishes the stop from the affricate. I do not commit myself here to the feature(s) required to distin-guish /p/ from /pf/ and /k/ from /kx/, but I assume that it is a (minor) place feature.Footnote 2

In this paper, I focus on the prediction in 4b. The reason 4b follows from the Strident Stop Representation is that in this model, the natural class of affricates and fricatives can only be captured if they share [+cont], but the representation in 3a has no [+cont] component. Neither 1a nor 2a makes the same prediction. In fact, both models claim that affricates and fricatives should form a natural class because both sets of sounds are [+cont].Footnote 3

The correctness of 4b is far from obvious; in fact, some superficial evidence suggests that it is wrong. For example, even many proponents of the Strident Stop Representation recognize the fact that sounds such as /s z / (fricatives) and /ts dz / (affricates) can pattern together in many languages (for example, LaCharité 1993, Clements 1999, Kehrein 2002). LaCharité (1993) demonstrates that with respect to Chumash Sibilant Harmony, Modern Yucatec Maya Morpheme Structure Constraints, Basque Stop Voicing, and Basque Stop Palatalization, fricatives such as /s/ and affricates such as /ts/ constitute a natural class. However, LaCharité (1993:7) correctly observes that in all these cases, affricates pattern only with sibilant fricatives: “Notably absent are rules that show affricates patterning with a full range of fricatives, including those that are not sibilant.” It is for this reason that she refers to the patterning of affricates together with fricatives as a “pseudoproperty” of affricates. A similar assessment of prediction 4b is provided by Clements (1999) and Kehrein (2002:46–57).

In the present article, I argue that the patterning of affricates in Cimbrian German requires reference to the natural class of affricates and fricatives, contrary to prediction 4b. In contrast to the evidence discussed by La Charité (1993), Clements (1999), and Kehrein (2002), the data to be discussed below involve both strident and nonstrident affricates patterning together with strident and nonstrident fricatives. Hence, the patterning together of affricates and fricatives in Cimbrian German cannot be analyzed as the natural class of [+stri] sounds.

3. Affricates in Cimbrian German

3.1 Overview of the Dialect

Cimbrian (German Zimbrisch) is an Upper German dialect spoken in northeastern Italy and derived historically from a Southern Bavarian dialect. According to the Ethnologue, there are approximately 2,300 native speakers of Cimbrian. The data to be discussed below have been drawn primarily from Tyroller 2003, which is a detailed description of the variety of Cimbrian spoken in the municipality of Lusern, in the province of Trento. I refer henceforth to this dialect as CLG (Cimbrian-Lusern German). Three additional sources for Cimbrian phonology are Bacher 1905, Kranzmayer 1981, and an unpublished handout by Alber 2011. A useful Internet reference for Cimbrian German dialects (including CLG), with a brief grammatical description and recordings, is available at http://titus.uni-frankfurt.de/didact/zimbr/cimbrian.htm.

Table 1 contains all CLG affricate, stop, and fricative phonemes. Tyroller (2003:62–71) provides numerous minimal or near minimal pairs demonstrating that the eighteen obstruents in table 1 are phonemes of CLG. Bacher (1905:161–165) provides some discussion of the phonetics of these sounds.

Table 1.

Obstruent phonemes in CLG.

There are several significant differences between the inventories of obstruents in CLG and MSG. First, CLG has the velar affricate /kx/; second, CLG has no fricative ; third, CLG has a voiceless sibilant Tyroller transcribes as which he describes as a “retracted voiceless ” (zurückgezogenes stimmloses ; p. 42). Tyroller (2003:49) later refers to the sound as retroflex, which is the term I adopt here. On the basis of Tyroller's examples it appears that CLG corresponds to West Germanic +/s/ and CLG /s/—to West Germanic +/t/.Footnote 4 The historical development of those sounds in Cimbrian dialects is discussed in Bacher 1905 and Kranzmayer 1981.

In the following sections, I consider the structure of affricates in CLG in light of the models discussed in section 2. Since all analyses agree that affricates are [−cont] (or the equivalent privative feature [stop]), I do not discuss evidence that the affricates in table 1 bear that feature. Instead, in sections 3.2 and 3.3 I concentrate solely on prediction 4b.Footnote 5 Following the studies cited in section 2, I assume that affricates in CLG, as in all languages, are, by definition, monosegmental.

3.2 The Verbal Allomorphs

According to Tyroller's (2003:90ff.) detailed description of the verbal morphology of CLG, there are five realizations of the infinitive morpheme: The same allomorphs also occur in the first and third person singular indicative verb forms. Bacher's (2005:193–197) material on verb conjugations in Cimbrian refers to these realizations as well. It is clear from both of these sources that the contexts for the verbal allomorphs are—for the most part—phonological because the same contexts account for the allomorphs of all other inflectional suffixes that end in a nasal (see section 3.3). In the following paragraphs, I describe the relevant data on CLG verbs and posit an analysis with (linear) rules. An alternative treatment in a different theoretical framework (for example, Optimality Theory) is possible, although I contend that any such alternative would need to capture formally the fact that affricates and fricatives pattern together.

The data in 5a–d illustrate that the [−әn] allomorph occurs after a verbal stem ending in a nasal, that is, [m n ŋ] or [g], respectively. It is shown in 6 below that [−ŋ] (and not [−әn]) occurs after a monosyllabic stem that ends in [g]. I, therefore, assume that [−әn] occurs in 5d only after [g] in a disyllabic verbal stem. Unless otherwise noted, the phonetic transcriptions in 5 and below have been taken directly from Tyroller 2003. Here and elsewhere, I present the MSG orthography in the first column and the CLG pronunciation in the second column.

  1. (5)

The examples in 6 show the distribution of −n [−n], −m [−m] and −ng [−ŋ]. The [n] suffix occurs after a vowel, as in 6a, or a liquid, as in 6b,c. In 6d–i, I show that the nasal suffix is homorganic with a stem-final voiceless or voiced stop. A process I refer to below as Nasal Place Assimilation (see 7a) ensures that the nasal suffix surfaces as [m] after /p/, as [n] after /t/, and as [ŋ] after /k/. The first member of each pair in 6g–i, that is, the infinitive form, has no voiced stop in the surface representation, but the second member of each pair illustrates that the voiced stop occurs in other members of the verbal paradigms. On the basis of these alternations, I hold that the underlying forms of the verb stems in 6g–i end in /b d g/. The −n (that is, /−n/) in those examples subsequently undergoes Nasal Place Assimilation, and the voiced stop then deletes by a process I refer to as Voiced Stop Deletion (see 7b).Footnote 6

  1. (6)

  2. (7)

The examples in 8 illustrate that the allomorph −an (that is, ) occurs after a fricative, as in 8a–g, or after an affricate, as in 8h–k. There is a wealth of data presented in Tyroller 2003 and Bacher 1905 for verbs with stems ending in all of the fricatives and affricates in 8. From Bacher 1905:172 it is also clear that only occurs after affricates and fricatives. The only gap is [dӡ].

  1. (8)

Note that the fricatives in 8a–g include both strident, that is, , as well as nonstrident sounds, that is, [f v x]. The same point holds for the affricates in 8h–k: surfaces after the strident affricates and after the nonstrident affricates [pf kx]. Italian loanwords with occurring after a fricative or affricate abound, for example, ankommen ‘arrive', compare Trentino rivàr.Footnote 7

As in Standard German (Wiese 1996), I hold that the infinitive ending of CLG is underlyingly /-n/ and that schwa is epenthesized before that consonant if it cannot be syllabified (C').Footnote 8 Schwa Epenthesis has been stated linearly in 9. A rule I posit below in 10 converts the epenthetic schwa to after an affricate or a fricative.

  1. (9)

A comment is in order concerning the context for Schwa Epenthesis. The disjunction in the left context ensures that schwa is inserted either after a nasal, as in 5a–c, or after an affricate or a fricative, as in 8. Note that the latter context is captured with the two features [−son, +cont]—a grouping that is only possible given the Contour Segment Representation in 1a or the Complex Segment Representation in 1b, but not the Strident Stop Representation in 3a. An analysis which collapses the nasal context and the affricate-fricative contexts into one is not possible under any approach to distinctive features known to me. One might suspect that if nasals are [+cont] (see Mielke 2002), then the two features [+cons, +cont] would do the job, but this is not the case because those features also describe /r/, which is a [+cont] sound.Footnote 9

The examples in 8 tell us that the allomorph surfaces after an affricate or a fricative and before a nasal. I capture the fact that the surface vowel is and not ә with the following process, which is implied in Tyroller's (2003:126) comment that there is a change from [әn] to [ n] after an affricate or a fricative:Footnote 10

  1. (10)

The rule in 10 is important for the present analysis because it refers to affricates and fricatives together as [−son, +cont], contrary to the prediction made by 4b. Significantly, the natural class of affricates and fricatives also subsumes nonstrident sounds, that is, [pf f x]. Note that the left context for Vowel Lowering is compatible with both the Contour Segment Representation in 1a and the Complex Segment Representation in 2a.Footnote 11 Five representative examples from the data sets provided above are illustrated in the following derivation:

  1. (11)

The derivations in 11 account for the distribution of the −an allomorphs in verbs as described in the original sources for CLG. As I demonstrate in the following section, the same analysis also holds for all nonverbal suffixes ending in a nasal.

3.3 Other Allomorphs

According to the description of CLG noun declensions in Tyroller 2003:126–127, there are five realizations of the plural morpheme for masculine and feminine nouns that are identical in shape to the realizations of the infinitive morpheme discussed in the previous section, that is, −en [−әn], −n [−n], −m [−m], −ng [−ŋ], and −an (see also Bacher 1905). The distribution of these nominal allomorphs is identical to the distribution of the homophonous verbal allomorphs discussed in section 3.2. Similar generalizations hold for surface allomorphs in adjective endings with a suffix-final nasal, demonstrative pronouns, and suffixes involved in deriving verbs from adjectives or nouns. As in section 3.2, we see below that it is the distribution of that provides evidence against the prediction in 4b because only occurs after fricatives and affricates.

The realization −en [−әn] of the plural allomorph surfaces after a noun stem ending in a nasal according to Tyroller (2003:126), although the only examples provided in that work are ones in which the nasal is [n]. The CLG noun [lαm] is a loan word (< Italian lama ‘blade’).

  1. (12)

In 13, one can observe that [−n] surfaces after liquids, as in 13a,b, while [−n], [−m], and [−ŋ] surface after homorganic stops, as in 13c–h. In the latter data set, /−n/ takes on the place features of a preceding voiceless stops [p t k] by Nasal Place Assimilation. In 13f–h, the nasal suffix surfaces as [m], [n], or [ŋ] after an underlying /b/, /d/, or /g/, respectively. In examples like these, Nasal Place Assimilation and Voiced Stop Deletion apply. I do not present the entire phonetic transcription for 13a,b because it is not given in Tyroller 2003.

  1. (13)

The most important set of data is presented in 14. Here −an surfaces after fricatives, that is, in 14a–f, or after affricates, that is, in 14g–j. (Examples 14a, d–g, and i–j are all from Bacher 1905). I have not found a masculine or feminine plural after a stem ending in [z]. I assume that this gap is accidental because the infinitive morpheme discussed in section 3.2 consistently surfaces as after fricatives and affricates, including [z].

  1. (14)

The data in 12–14 show that the plural allomorphs [−әn], [−n], [−m], [−ŋ], and have the same distribution as the homophonous verbal allomorphs discussed in section 3.2.

An identical alternation can be observed for adjectival suffixes. Tyroller (2003:143) states clearly that occurs after fricatives and affricates, [−әn] after nasals, [−n] after vowels, liquids, and (alveolar) stops. In 15, I provide representative examples for these realizations in the dative/accusative singular of the weak declension (Tyroller 2003: 144–145). (Although Tyroller does not provide examples of adjective stems ending in an affricate, it is clear from that source that the allomorph occurs in that context in 15, as well as 16 and 17).

  1. (15)

As shown in 16, the same generalization that governs the distribution of [−әn], [−n], and holds for the various realizations of the weak plural declension and for the dative/accusative singular of the strong declension.

  1. (16)

The dative singular endings for the demonstrative suffix reveal the same pattern of allomorphy.

  1. (17)

The final data set (Tyroller 2003:190–192) illustrates that suffixes ending in a nasal can convert nouns into adjectives, as in 18a,b, or they can convert nouns and adjectives into verbs, as in 18c. Observe that [−әn] occurs after a nasal in 18a, while occurs after fricatives or affricates in 18b,c.

  1. (18)

In sum, the distribution of regardless of the lexical category of the base word, is determined entirely by the phonology: [−әn] occurs if the final segment of the stem is an affricate or fricative, [−әn] after a nasal, [−n] after a vowel or liquid, and [−n]/[−m]/[−ŋ] after a homorganic stop. The analysis proposed in the preceding section for verbs can easily be extended to the data in 12–18.

4. Conclusion and Open Questions

In the present article, I have argued that affricates and fricatives in the Cimbrian dialect of German (CLG) pattern together in the context of two phonological rules. This finding is important because it contradicts the claim made in previous studies that affricates and fricatives do not constitute a natural class (recall 4b). The conclusion is that affricates in CLG must have a [+cont] component and that the natural class of affricates and fricatives is therefore characterized as [−son, +cont]. I remain agnostic concerning the complete featural representation for affricates in CLG. My conclusion is that either the Contour Segment Representation in 1a or the Complex Segment Representation in 2a is compatible with the material presented above. In the future, one might want to investigate additional data from CLG—or other varieties of Cimbrian German—involving affricates in order to determine which of the two representations does the most justice to the facts.

Several fundamental questions remain open for further study. First, it would be a worthwhile endeavor to search for additional evidence from CLG that would confirm the findings of the present study. This dialect could conceivably have additional rules that treat affricates and fricatives as a unit, for example, if only affricates and fricatives (but not stops) undergo a certain process. It is also possible that fricatives and affricates function as a natural class in CLG if these and only these sounds are subject to certain phonological rules or phonotactic constraints.

Other open questions have a wider scope. For example, a very obvious question to consider in future work is whether or not there are other languages (or dialects of German), in which affricates and fricatives pattern together. One of the reasons why further research on German dialects like CLG is important is that they have nonstrident affricates such as /pf/ and /kx/, as pointed out above. Some of the studies referred to in section 2 have argued that affricates cannot and should not be treated as [+cont], but it remains to be seen if all of these case studies have been successfully reanalyzed from the perspective of the Strident Stop Representation.

Several unrelated issues clearly worthy of future study were mentioned in the course of my analysis. For example, one could investigate the phonology and phonetics of the CLG vowel (and [ә]) as well as the nature of the proposed process in 10 as a lowering or a retraction. Finally, there needs to be a thorough investigation of the (place) features distinguishing the sibilants of CLG, especially and .

Footnotes

* I am indebted to two referees for their critical comments on an earlier incarnation of this article. All disclaimers apply.

1 Lombardi stresses that in her model, [stop] and [cont] are unordered throughout the phonology, but that they are ordered phonetically. This distinction between levels predicts that edge effects should, in principle, be possible for phonetic processes. Lin (2005) has taken the opposing position in arguing that the affricates of the Arawakan language Piro (spoken in Peru) require a representation such as 1a in the lexical phonology and such as 2a in the postlexical phonology. The reason is that the affricates in Piro display edge effects in the lexical phonology and antiedge effects in the postlexical phonology.

2 The sounds I transcribe as /k/ and /kx/ may differ in that the former is velar and the latter is uvular (or the other way around). Whether or not this is always the case in languages with /k/ and /kx/ is a question I leave for further study. As I point out in note 5, this question is germane because Cimbrian German has both sounds.

3 There are a number of alternate representations for affricates which I have not discussed here, for instance, Clements & Keyser 1983, van de Weijer 1994, and Shafer 1995. All of these studies agree that affricates are [−cont] and [+cont], although individual models differ with respect to the relationship between these and other features. Since the aforementioned studies see affricates as [+cont], none of them makes the prediction in 4b.

4 Tyroller (2003:67) lists [z] as an allophone of /s/. Alber (2011) similarly argues that all voiced fricatives of Cimbrian dialects (including CLG) are allophones of the corresponding voiceless fricatives: The former occur word initially before a vowel and intervocalically if the first vowel is long, while the latter occur in word-final position after a vowel and intervocalically if the first vowel is short. I include /z/ as a phoneme because this sound contrasts with other sibilants in word-initial position before a vowel and intervocalically, for example, ‘saw’ versus ‘scythe’ (Tyroller 2003:67). The confusion around the phonemic status of sounds such as [z] is not important for the present analysis as it only addresses the behavior of sibilants in stem-final position (see sections 3.2 and 3.3). In stem-final position, labiodentals and alveolar fricatives are predictably voiceless after short vowels, and voiced after long vowels and sonorant consonants. Other fricatives, that is, , are invariably voiceless in stem-final position (and in all other contexts). Tyroller (2003) interprets as an allophone of /s/ and , but it is difficult to see how this is possible given the many examples in his descriptive grammar in which contrasts with [s] and , for example, miss ‘Mass’ versus misch ‘mix’ (Tyroller 2003:69).

5 A potential challenge for the Strident Stop Representation-in particular, prediction 4a-concerns the exact phonetic realization of the CLG nonstrident affricate [kx]. While Tyroller (2003:44) describes both the stop [k] and the affricate [kx] as velar, it is also conceivable that [kx] is uvular and [k] is velar, in which case prediction 4a could, in principle, be upheld. In light of the paucity of phonetic evidence, I leave this question open for further study.

6 I assume that [−m], [−n], and [ŋ] in 6d-f are syllabic on the surface, although Tyroller (2003) does not indicate this in his phonetic transcriptions. One can only speculate whether or not the [−n] in 6a-c is syllabic (as in MSG).

7 While copious examples can be found for after the dorsal affricate [kx] (see 8k), Tyroller is inconsistent concerning how this and similar suffixes (see section 3.3) are realized in that context. On the one hand, we have evidence that surfaces after [kx], but on the other hand, Tyroller notes that the [−ŋ] allomorph is possible in some words (see, for example, pp. 46, 93). For example, he transcribes the verb grab as In the 200-page dictionary of Cimbrian in Bacher 1905:211–432, it is very easy to find nouns whose stem ends in [kx] followed by after [kx] in addition to the ones listed in 8k, but I have not found any examples in that source with in the same context. One could speculate that Tyroller's speakers showed free variation, that is, both and are possible, while Bacher's speakers were much more consistent. This is clearly an issue that needs to be left open for further study.

8 A consonant ‘A’ cannot be syllabified into the coda of the preceding syllable ending in ‘B’ if A is more sonorous than B. For example, the underlying /−n/ suffix in 8 follows a stem ending in a fricative. Since the former is more sonorous than the latter, Schwa Epenthesis will insert [ә] between the two sounds. Note that the [ә] in 5d cannot be inserted by Schwa Epenthesis because the final segment of the stem (that is, /g/) does not satisfy either of the two parts of the disjunction in its left context. It should be clear that these data pose a problem for any treatment of CLG. I hold that examples like the ones in 5d require a lexically specified suffix /−әn/.

9 An alternative to Schwa Epenthesis is a treatment which analyzes [−n], ә, and [−әn] as suppletive allomorphs (see Booij 1998 and Paster 2006 for this type of analysis in other languages). Under such an analysis, 9 would not be necessary given that there would be subcategorization frames specifying the contexts for [−n], , and [−әn]. (In a suppletive treatment of CLG, the natural class of affricates and fricatives is expressed in the subcategorization frame for ). The argument against treating the verbal allomorphs of CLG as suppletive is that the same type of subcategorization would be required for all other CLG suffixes that end in a nasal (see section 3.3).

10 In contrast to MSG, in CLG is not an allophone of /r/. It is not difficult to find examples in Tyroller 2003 in which [ә] contrasts with and other vowels, for example, dann ‘then ’, Honig ‘honey’, schön ‘beautiful’, Boden [po:n] ‘floor’. I refer to the rule in 10 as a lowering because I am assuming that [ә] is central and is (upper) low, although it needs to be clarified what type of vowel is from the phonological perspective. One possibility is that is back (that is, [+back]) and /ә/ is central (that is, [−back]), in which case it would make more sense to refer to 10 as a rule of “retraction.” Some independent evidence for the latter interpretation is that CLG appears to have a process that retracts the only other central vowel phoneme, that is, /a/, before a nasal. See the discussion in Tyroller 2003:30, in which central [a] and back [α] are allophones, whereby the latter only occurs before a nasal or before a historical nasal that was later deleted. The interpretation of 10 as lowering or retraction is immaterial for the present analysis.

11 An anonymous reviewer points out that the natural class of affricates and fricatives need not require structures with [+cont] as in 1a or 2a. Alternatively, there could be a place of articulation feature working across major articulators which does the job equally well. For example, (recall section 2) could describe the sounds after which occurs, that is, . An obvious problem with this approach is that (recall 8e) is usually considered to be [+distr], and yet occurs after as well. By contrast, my treatment, which treats and as [+cont], has no such drawback. I argue that the burden of proof now lies on the shoulders of linguists who want to argue in favor of [distr] or some other feature.

References

REFERENCES

Alber, Birgit. 2011. Voicing contrasts in Mòcheno and Cimbro. Paper presented at the CUNY Conference on the Phonology of Endangered Languages, January 12–14, 2011. New York, NY: CUNY.Google Scholar
Bacher, Joseph 1905. Die deutsche Sprachinsel Lusern. Quellen und Forschungen zur Geschichte, Literatur und Sprache Österreichs und seiner Kronländer X. Innsbruck: Verlag der Wagner'schen Universitäts-Buchhandlung.Google Scholar
Booij, Geert 1998. Phonological output constraints in morphology. Phonology and morphology of the Germanic languages, ed. by Kehrein, Wolfgang & Wiese, Richard, 143–163. Tübingen, Germany: Niemeyer.CrossRefGoogle Scholar
Clements, G. N. 1999. Affricates as noncontoured stops. Item order in language and speech, ed. by Fujimura, Osamu, Joseph, Brian D., & Palek, Bohumil, 271–299. Prague: Charles University Press.Google Scholar
Clements, G. N., & Keyser, Samuel Jay. 1983. CV Phonology. A generative theory of the syllable. Cambridge, MA: MIT Press.Google Scholar
Hall, Tracy Alan. 2004. Assibilation in Modern German. Lingua 114. 1035–1062.CrossRefGoogle Scholar
Hualde, Jose I. 1988. Affricates are not contour segments. Proceedings from the 7th West Coast Conference on Formal Linguistics, ed. by Borer, Hagit, 143–157. Stanford, CA: CSLI Publications.Google Scholar
Jakobson, Roman, Fant, Gunnar, & Halle, Morris 1952. Preliminaries to speech analysis. Cambridge, MA: The MIT Press [sixth printing, 1965].Google Scholar
Kehrein, Wolfgang. 2002. Phonological representation and phonetic phrasing: Affricates and laryngeals. Tübingen: Niemeyer.CrossRefGoogle Scholar
Kim, Hyunsoon. 2001. A phonetically based account of phonological stop assibilation. Phonology 18. 81–108.CrossRefGoogle Scholar
Kranzmayer, Eberhard. 1981. Laut- und Flexionslehre der deutschen zimbrischen Mundart. Beiträge zur Sprachinselforschung Band I. Vienna: VWGÖ.Google Scholar
LaCharité, Darlene. 1993. The internal structure of affricates. Ottawa: University of Ottawa dissertation.Google Scholar
Lin, Yen-Hwei. 2005. Piro affricates: Phonological edge effects and phonetic anti-edge effects. The internal organization of phonological segments, ed. by van Oostendorp, Marc & van de Weijer, Jeroen, 121–152. Berlin: Mouton.CrossRefGoogle Scholar
Lin, Yen-Hwei. 2011. Affricates. The Blackwell companion to phonology, ed. by van Oostendorp, Marc, Ewen, Colin J., Hume, Elizabeth, & Rice, Keren. (Oxford: Blackwell reference online http://www.companiontophonology.com/subscriber/tocnode?id=g9781405184236_chunk_g978140518423618. Accessed July 5, 2011.)Google Scholar
Lombardi, Linda. 1990. The nonlinear organization of the affricate. Natural Language and Linguistic Theory 8. 375–425.CrossRefGoogle Scholar
Mielke, Jeff. 2005. Ambivalence and ambiguity in laterals and nasals. Phonology 22. 169–203.CrossRefGoogle Scholar
Paster, Mary. 2006. Phonological conditions on affixation. Berkeley, CA: University of California at Berkeley dissertation.Google Scholar
Rubach, Jerzy. 1994. Affricates as strident stops in Polish. Linguistic Inquiry 25. 119–143.Google Scholar
Sagey, Elizabeth. 1986. The representation of features and relations in phonology. Cambridge, MA: MIT dissertation.Google Scholar
Shafer, Robin. 1995. Headedness in the representation of affricates. Linguistic Review 12. 61–87.Google Scholar
Tyroller, Hans. 2003. Grammatische Beschreibung des Zimbrischen von Lusern. Wiesbaden: Franz Steiner Verlag.Google Scholar
Van de Weijer, Jeroen. 1994. Segmental structure and complex segments. Leiden: Rijksuniversiteit te Leiden dissertation.Google Scholar
Wiese, Richard. 1996. The Phonology of German. Oxford: Clarendon Press.Google Scholar
Figure 0

Table 1.Obstruent phonemes in CLG.