Morphology (2007) 17:151–178 DOI 10.1007/s11525-007-9112-z

ORIGINAL PAPER

Inflectional paradigms have a base: evidence from s-Dissimilation in Southern

T. A. Hall Æ John H. G. Scott

Received: 7 March 2007/Accepted: 4 June 2007/Published online: 15 September 2007 © Springer Science+Business Media B.V. 2007

Abstract In many varieties of Southern German the contrast between /s/ and /∫/ is neutralized to [∫] before /p t/ anywhere within a word (e.g. Post [po∫t] ÔmailÕ), but neutralization does not occur before inflectional suffixes (e.e. ku¨ss-t [kyst] ‘kiss (3 SG)’). It will be argued that the underapplication of neutralization before inflectional suffixes is an example of a Paradigm Uniformity effect: Neutralization is blocked from applying to the final /s/ of a stem so that it will retain a constant shape in a paradigm. Underapplication in examples like [kyst] follows from a requirement that the stem in a derived word be identical to the unaffixed base. By contrast, the German data will be shown to be problematic for the Optimal Paradigms model (McCarthy 2005), since this approach does not allow for a base in inflectional paradigms.

Keywords Paradigm uniformity Æ Base identity Æ Optimality Theory Æ German Æ Swabian Æ Realize Morpheme Æ Homophony

T. A. Hall (&) Æ John H. G. Scott Department of Germanic Studies, Indiana University, Ballantine Hall 644, 1020 Kirkwood Avenue, Bloomington, IN 47405-7103, USA e-mail: [email protected]

J. H. G. Scott [email protected]

123 152 T. A. Hall, John H. G. Scott

1 Introduction

In many southern German dialects (e.g. Swabian) the contrast between /s/ and /∫/ is neutralized to [∫] before /p t/ anywhere within a word. Representative examples illustrating the process we refer to as s-Dissimilation are presented in (1a). s-Dissimilation is systematically blocked before (inflectional) suffixes, as in (1b). These data have been collected from a Swabian speaker from Donau- eschingen:

(1) Examples illustrating s-Dissimilation (in a) and its blockage (in b):

a a. stark [∫t c k] ‘strong’ Post [po∫t] ‘mail’ b. ku¨ ss-t [kyst] ‘kiss (3 SG)’

In the terminology of rule-based phonology s-Dissimilation underapplies in (1b). This means that the structural description of the rule is met but that the process nevertheless fails to go into effect. We argue that the underapplication of s-Dissimilation in (1b) derives an explanation based on the notion of Paradigm Uniformity (e.g. Raffelsiefen 1995; Kenstowicz 1996; Benua 1997; Steriade 2000; Raffelsiefen 2004 and the contributions in Downing et al. 2005). One of the central ideas of Paradigm Uniformity is that a phonological process (e.g. s-Dissimilation) can be blocked from applying or favored in applying to increase the identity of morphologi- cally related words. In Optimality Theory (OT; Prince and Smolensky 2004; Kager 1999) Para- digm Uniformity is commonly captured by positing specific Output-Output (O-O) constraints, which by definition demand identity to morphologically related words. Although O-O constraints are frequently employed in the liter- ature there is little agreement on how they should be evaluated. In derivational morphology there appears to be agreement that O-O constraints compare affixed forms with an unaffixed base (e.g. Raffelsiefen 1996, 2004; McCarthy 2005), but there is disagreement in the literature concerning whether or not this approach also holds for inflection. Some linguists have argued that inflectional paradigms have a base to which the remaining forms in the paradigms must be faithful (e.g. Kenstowicz 1996; Albright 2004), while others hold that inflec- tional paradigms do not require reference to a base (e.g. the Optimal Paradigms (OP) model of McCarthy 2005). Thus, in the latter approach McCarthy for- malizes O-O constraints (called OP constraints in his model) in such a way that each member of an inflectional paradigm must match every other member but without a base. The OP model makes a very clear prediction concerning the type of PU effects which should be possible in natural language. If a canonical, phono- logical pattern requires the general ranking in (2a), then McCarthy shows that the regular phonology can spread to the rest of the paradigm (i.e. the regular rule overapplies), by means of a high-ranking OP constraint, as in (2b).

123 Inflectional paradigms have a base 153

(2) Paradigm Uniformity effects in the OP model: a. Canonical pattern: Markedness » I-O Faithfulness b. Paradigm Uniformity effects (overapplication): OP Faithfulness » Markedness » I-O Faithfulness

McCarthy (2005, p. 197) stresses that the ranking in (2b) always results in the overapplication paradigm being selected over the underapplication paradigm and that underapplication in inflectional paradigms is a situation that can only be possible if there is some high ranking (markedness) constraint which penalizes the overapplication paradigm fatally. We argue below that the data in (1) are problematic for McCarthy’s pre- diction. In particular, we show that for German it is not clear what the high ranking constraint would be which would allow the underapplication paradigm to be selected as optimal. We propose instead that the data in (1) follow if the O-O constraint is faithful to a base, contrary to the approach taken in McCarthy (2005). A second problem dealt with below is the realization of the second person singular in verbs whose stem ends in a sibilant . In the southern German dialects we discuss, the stem-final segment deletes, e.g. gru¨ssen ‘greet, 2 SG’[griE-∫] (from /griEs-∫/), in violation of the base-identity requirement. Our analysis therefore needs to answer two questions: First, why are base-identity effects sacrificed for segmental deletion? And second, why is the final segment of the stem deleted and not the sibilant of the suffix? We argue that faithful outputs like *[griEs∫] which would satisfy base-identity do not occur because German avoids a sequence of two sibilants and that the deletion of the suffix does not occur (*[griEs-]) because there is a strict requirement that every morpheme receive some phonological expo- nence (Kurisu 2001). Varieties of southern German in which the second person singular ending is [∫t] (and not [∫]) prohibit the deletion of the [∫] in the suffix because the output would be homophonous to another member of the same paradigm (Crosswhite 1999; Kenstowicz 2005). The present is structured as follows. In Sect. 2 we present the data from Swabian German — representative of the southern German dialects referred to above — which motivate s-Dissimilation and then give an analysis thereof. In Sect. 3 we introduce the data illustrating the underapplication of s-Dissimilation. Our analysis of s-Dissimilation in (1) as a process requiring a constraint ensuring base- identity is presented in Sect. 4. In Sect. 5 we discuss the OP model of McCarthy (2005) and show that it makes the incorrect predictions concerning the German data. In Sect. 6 we discuss the case of the second person singular in verbal para- digms. Sect. 7 presents our conclusions and points to directions for further research.

2 s-Dissimilation: the canonical pattern

The two coronal /s/ and /∫/ are uncontroversially phonemes with similar distribution in both and Swabian German (SwG), as

123 154 T. A. Hall, John H. G. Scott illustrated in (3) below. We see here that these two fricatives contrast word- initially before a vowel (in a), word-finally after a vowel (in b) and intervo- calically (in c).1

(3) Contrast between [s] and [∫] in Standard German and SwG: Standard SwG a. Saal [zaːl] [saːl] ‘hall’ [Frey 1975: 14] Schale [∫aːlE][∫aːl] ‘bowl’ [Frey 1975: 14] b. Bus [bʊs] [bus] ‘bus’ [Frey 1975: 34] Rausch [Rau∫] [rEu∫] ‘intoxication’ [Frey 1975: 11] c. Massen [masEn] [masE] ‘masses’ [Frey 1975: 14] Maschen [ma∫En] [ma∫E] ‘meshes’ [Frey 1975: 14]

It should be noted that the opposition between /s/ and /z/ in Standard German is neutralized to [s] everywhere in SwG. The postalveolar fricatives /∫ Z/ have been neutralized in the same manner to [∫]. The contrast between /s/ and /∫/ is neutralized to [∫] before /p t/ in both varieties of German (see 4a–b). In Standard German this neutralization occurs word-initially but in SwG it occurs throughout the word. In (4c) we can observe that no neutralization takes place before /k/ in either dialect.2

(4) Contrast between [s] and [∫] is neutralized to [∫] before [tp]: Standard SwG a. Speck [∫pek] [∫pek] ‘bacon’ Speise [∫paizE][∫pais] ‘food’ Knospe [knOspE] [kno∫pE] ‘bud’ Haspel [haspEl] [ha∫pEl] ‘hasp’ Wespe [vespE] [ve∫p] ‘wasp’ [Frey 1975, p. 56]

b. stark [∫taRk] [∫Oak] ‘strong’

Staat [∫taːt] [∫taːt] ‘country’

a Fenster [fenst ]a [fen∫t ] ‘window’ Post [pOst] [po∫t] ‘mail’

1 Our primary source of data for SwG is Frey (1975), who describes the variety of that dialect spoken in . The relevant facts are consistent with those in other varieties of SwG and other southern German dialects to be discussed below. The examples below not from Frey (1975) or another published source were provided by Vera [Hagen] Hausherr from Donaueschingen. 2 With the exception of nonintegrated loan words like Spha¨re [sfeːRE] ‘sphere’ and Sphinx [sfɪnks] ⌢ ‘sphinx’, /s/ cannot precede fricatives. The only /s/ plus sequences involves [sts], e.g. Szene ⌢ [stseːnE] ‘scene’. Jutz (1931, p. 199) writes that neutralization takes place before /k/ in some varieties of southern German. One example (not given by Jutz) is the Swiss dialect described by Wipf (1910), in which the facts seem to be the same as in SwG with the exception that /s/ also shifts to [∫] before /k/ (p. 87). As noted by Wipf, these examples are loan words, since all original /sk/ clusters in native German words shifted this sequence to [∫]; e.g. German Schiff [∫ɪf] ‘ship’ (< Old High German scif).

123 Inflectional paradigms have a base 155

c. Skelett [skelet] [skElet] ‘skeleton’ bru¨ sk [bRYsk] [bRysk] ‘brusque’ Kiosk [kiːOsk] [kiOsk] ‘kiosk’

The pattern of neutralization in (4a–b) is referred to in the literature as s-Dissimilation (Wiese 1991; Hall 1992; Alber 2001; Scott 2006). The reason why neutralization involves a dissimilation is not crucial to our analysis, but it will be discussed below. The within-word context for s-Dissimilation is typical not only for SwG, but for most dialects spoken in southwest (including parts of Switzerland and Austria). See Jutz (1931, p 199), Schirmunski (1962, p. 361–362) and Alber (2001, p. 26) for some general discussion. We concentrate below on the SwG pattern, although we also make some reference to dialects outside this region (which we refer to collectively as ‘southern German’). We follow the authors cited above who see s-Dissimilation as an active process because it also applies to (recent) loan words in both SwG and Standard German. Representative examples of loanwords which undergo s-Dissimilation are provided in (5):

(5) s-Dissimilation in loan words:

Stil [∫tiːl] ‘style’ [Wiese 1991, p. 129] Spezies [∫peː⌢tsjEs] ‘species’ [Wiese 1991, p. 129] Protest [prote∫t] ‘protest’ [Russ 1982, p. 76]

We now posit the following provisional rule:

(6) Provisional rule of s-Dissimilation in SwG: /s/ fi [∫] / __ [p t ]

The feature which changes in s-Dissimilation has been argued to be [high] as opposed to [anterior] (see Hall 1992; Wiese 1996; Alber 2001). Various models of feature geometry have proposed that [high] is independent of all articulators (i.e. [LABIAL], [CORONAL] and [DORSAL]); see, for example, Lahiri and Evers (1991). The relevant place features for German consonants we are assuming are presented in (7):

(7) Place features for German consonants

p b pf m f v t d ts s n l r [LAB] [COR] [DOR] [high] – – + + –

As can be seen in (7), the segments /p t/ are [–high], while /k/ is [+high]. s-Dissimilation in SwG can now be formalized featurally as in (8):

123 156 T. A. Hall, John H. G. Scott

(8) s-Dissimilation rule in SwG: /s/ fi [+high] / __ [–sonorant, –continuant, –strident, –high] In terms of features, s-Dissimilation only applies to a [–sonorant, +continuant, ⌢ CORONAL] segment. Since the coronal affricate /ts/ is not [+continuant], these features ensure that the rule only affects /s/. The analysis of as strident stops assumed here is defended by a number of phonologists, e.g. LaCharité (1993), Rubach (1994), Clements (1999), Kim (2001), and Kehrein (2002), following Ja- kobson et al. (1952). We have written s-Dissimilation in such a way that the process does not apply before sonorants because the sonorant environment is different from the envi- ronment before /p t/. Consider first the regular neutralization of /s/ to [∫] in both varieties of German before sonorant consonants (including [v]), as illustrated in (9).

(9) Neutralization of /s/ and /∫/ to [∫] before sonorant consonants (including [v]): Standard/SwG schreiben [∫RaibEn] ‘to write’ Schlange [∫laNE] ‘snake’ Schmuck [∫mʊk] ‘jewelry’ Schnee [∫neː] ‘snow’ ⌢ schwarz [∫vaRts] ‘black’ However, in contrast to the loan words in (5) with initial /sp st/, loan words with initial /sn sm sl/ do not undergo s-Dissimilation, e.g. Snob [snOp] ‘snob’ (*[∫nOp]), Slalom [slaːlOm] ‘slalom’ (*[∫laːlOm]). This generalization also extends to word- internal clusters, e.g. Kosmos [kOsmoːs] ‘cosmos’ (*[kO∫moːs]), Oslo [Oslo] ‘Oslo’ (*[O∫lo]). Although the rule we posited above in (8) correctly applies before /p t/ only, we do not have an explanation for why s-Dissimilation fails to apply productively before sonorants. It is also worth noting that the literature referred to above does not address this issue either, so this is a question that must remain open for further study. The important point is that the analysis we posit below of underapplication in Sect. 4 does not crucially depend on an answer to the question because underapplication occurs precisely where s-Dissimilation appliesproductively,namelybefore/t/. Following Alber (2001), we see an OT analysis of s-Dissimilation as a response to the conflict between the specific OCP constraint in (10a) and the faithfulness constraint in (10b). This conflict is illustrated in the tableau in (11).3

3 Alber (2001) observes several potential problems with constraint (10a). First, it is too general ⌢ because it also penalizes sequences like /pl/ (e.g. Platz [plats] ‘seat’), /ps/ (e.g. Gips [ɡɪps] ‘plaster’) and /sn sm sl/ (e.g. Kosmos [kOsmoːs] ‘cosmos’). Second, there are also dialects of southern German in which s-Dissimilation applies productively to /s/ in /sk/ clusters. The first problem is addressed in Hall (2007) but cannot be repeated here for reasons of space. The second point referred to above means that for speakers of these dialects the neutralization of /s/ before /p t k/ is not a dissimilatory change, in which case OCP-[high] would have to be substituted with a markedness constraint which penalizes [s] plus any stop. Note that an answer to these questions is not a prerequisite to the main point of the paper, which is an explanation of the underapplication of s-Dissimilation before /t/.

123 Inflectional paradigms have a base 157

(10) Constraints necessary for the canonical pattern: a. OCP-[high]: [ahigh] [ahigh] is disallowed b. IDENT -[high]: corresponding segments are identical in the feature [ high]

(11) s-Dissimilation for the example Last [la∫t] ‘burden’ in SwG:

/last/ OCP-[high] IDENT -[high] a. [last] *! b. [la t] *

Candidate (11a) violates OCP-[high] fatally because of the adjacent [s] and [t] and therefore candidate (11b) is selected as optimal.4 The analysis in (11) is in con- formance with Richness of the Base. Thus, a potential input form /la∫t/ would correctly surface as [la∫t], because its closest competitor [last] violates OCP-[high].

3 The context for blocking

The examples in (12) illustrate that s-Dissimilation does not apply if the /s/ and the following consonant are separated by a morpheme boundary. This gener- alization is true across a compound juncture, so if the /s/ ends the first member of the compound and the non-high consonant begins the second part (see 12a), then /s/ surfaces as [s].5 The same situation obtains in prefixed words in which the prefix ends in /s/ and a non-high consonant begins the stem (see 12b).6 The final set of examples (see 12c) demonstrates that s-Dissimilation is blocked from applying to a stem-final /s/ followed by an inflectional suffix (which can only be /t/). There do not appear to be any derivational suffixes beginning with /t/ (or /p/) which could potentially attach to a base ending in /s/ (see Frey 1975 for the Stuttgart variety of SwG and Fleischer and Barz 1995 for Standard German).7

4 Candidate (11a) has two possible representations; namely one in which [s] and [t] each have their own instantiation of [–high] and one in which [–high] is shared by both segments. Only the first representation (which corresponds to what we have in 11a) violates OCP-[high]. The second rep- resentation violates Crisp-Edge [Segment], a constraint penalizing linked structures across segments, which we omit below (Ito^ and Mester 1999, p. 208; Noske 1997, p. 225) 5 One occasionally encounters descriptions of specific dialects in which s-Dissimilation applies across the two parts of a compound. See, for example, Bertram’s (1937, p. 137) description of the dialect spoken in the Palatinate. We do not attempt to account for the domain of s-Dissimilation in such dialects. 6 Miss- is a common prefix. Frey (1975, p. 108) lists the uncommon, nonnative prefix dis-, e.g. ⌢ diskret [dɪskret] ‘discreet’ vs. Distanz [dɪ∫tants] ‘distance’, in which s-Dissimilation applies in the second example. We propose that speakers who apply s-Dissimilation in words like Distanz no longer perceive of dis- as being a prefix. 7 The Standard German derivational suffix –t, e.g. Fahr-t ‘drive (noun)’ (cf. fahr-en ‘drive (verb)’) is such a suffix, but it is not productive at all. More to the point, there are no stems ending in /s/ followed by this /t/ which could potentially undergo s-Dissimilation (Muthmann 1988).

123 158 T. A. Hall, John H. G. Scott

(12) s-Dissimilation blocked across a morpheme boundary in SwG: a. Kinderbrei [kendlesbrEi] ‘pap’ [Frey 1975: 120] b. miss-trauen [mistrEuE] ‘mistrust’ [Frey 1975: 107] c. pass-t [past] ‘fit (3 sg)’ [Frey 1975: 14] ge-wuss-t [gvist] ‘knew (PAST PART)’

According to our survey of the literature on German dialects the blockage of s-Dissimilation across morpheme boundaries as in (12) is true in virtually all of the regional varieties of German spoken within the geographical area mentioned in Sect. 2, although it needs to be stressed that many of these works only list the (12c) context and the ones in (12a–b) are not discussed. For other varieties of SwG showing the same pattern of blockage in (12) see Haag (1898, p. 37), who describes the dialects spoken in Schwenningen. See also Wipf (1910, p. 88) and Berger (1913, p. 131), who describe Waliser German (Wallis, Southern Switzerland) and St. Gallen German (Eastern Switzerland) respectively. According to Schatz (1897, p. 92) the same pattern holds in a variety of Bavarian German spoken in Western Austria. See also Tarral (1903, pp. 55, 105–114), who observes the same type of blockage in the Falkenberg dialect spoken in Lorraine (present day France, De´ partement Moselle). We do not consider the blockage of s-Dissimilation in (12a–b) to be par- ticularly remarkable, since segmental processes in many other languages (including German) are inhibited from applying in precisely those environ- ments. We account for (12a–b) by adopting the proposal made independently for German by other linguists (e.g. Booij 1985; Wiese 1996; Raffelsiefen 2000; Hall 2002) that stems like [kendles], [brEi], and [trEuE] are independent pho- nological words, and that s-Dissimilation only applies when the /s/ and the following non-high consonant belong to the same phonological word. From a formal point of view the analysis in (11) can capture the blockage of s-Dis- similation in contexts (12a–b) by modifying OCP-[high] so that the two instantiations of [ahigh] referred to in that constraint only count as an OCP- [high] violation if they both belong to the same phonological word. Given this prosodic restriction, the adjacent [st] and [sp] sequences in (12a–b) will surface faithfully. The prosodic solution described in the preceding paragraph makes sense for (12a-b), but it fails to account for under application in (12c). The reason is that in any analysis these inflectional suffixes must belong to the phonological word of the stems to which they attach. Thus, the question that needs to be answered is why s-Dissimilation underapplies in (12c).8

8 One could imagine a prosodic solution according to which an example like /pas-t/ is parsed

((pas)xt)CG, where `x' and `CG' denote `phonological word' and `clitic group' respectively, and s-Dissimilation does not apply to the /s/ because the trigger and target must belong to the same phono- logical word. Since there is no independent evidence at all for (a) analyzing /t/ as being outside of the phonological word of the stem and (b) postulating some additional prosodic category like the clitic group, we contend that the burden of proof lies on the shoulders of linguists who want to uphold this type of analysis.

123 Inflectional paradigms have a base 159

Given the examples in (3–6) and the additional examples in (12) one might hypothesize that s-Dissimilation only applies within a stem. The additional examples in (13) show that this domain is not sufficient to account for the entire range of data because s-Dissimilation also applies to the /s/ if it lies within the same suffix as the following consonant. There are no prefixes to our knowledge in which /s/ occurs to the immediate left of a non-high consonant.

(13) s-Dissimilation applies within a suffix: a. superlative suffix: doof-ste [doːf∫t] ‘most stupid’ [Frey 1975, p. 56] plump-ste [plomp∫t] ‘most awkward’ [Frey 1975, p. 57] b. second person singular suffix: stirb-st [∫tirp∫t] ‘die’ [Zinser 1933, p. 29]

In (13a–b) we can observe that the /s/ of the superlative suffix and the second person singular suffix surface as [∫] because it is followed by /t/.9 In Sect. 4 and 5 we concentrate on the underapplication of s-Dissimilation before inflectional suffixes as in (12d). One conceivable analysis of these data, which we do not adopt here but discuss for the sake of thoroughness, would be to posit different inputs for tauto- and heteromorphemic /st/ sequences such as those depicted in (14):

(14) a. b.

The structure in (14b) is similar to the ones proposed by Ní Chiosain (1991, p. 34) and Grijzenhout (1998, pp. 41–42) resulting from processes they call ‘‘coronal fusion’’ – a process which is intended to explain the blockage of other phonological processes in derived clusters. Given the input structures in (14) and the constraint in (15), the pattern of s-Dissimilation in (14a) and its blockage in (14b) would follow, as illustrated by the tableaux in (16) and (17).

(15) O-CONTIGUITY (‘No Intrusion’): The portion of the output standing in correspondence forms a contiguous string (McCarthy and Prince 1995, p. 123; Alber 2001, p. 13)

9 As we note in section 6, the second person singular suffix in some varieties of SwG is [∫t] (Bohnenberger 1928: 46, Zinser 1933: 28ff.), and in other ones it is [∫], which we analyze synch- ronically as /∫/ (e.g. Frey 1975). Our informant appears to pronounce the suffix as [∫] before a word beginning with a stop and as [∫t] before a word beginning with a vowel. Since we do not have a complete set of phonological and syntactic contexts in which deletion of [t] occurs, our analysis abstracts away from the [∫] [∫t] alternation. 

123 160 T. A. Hall, John H. G. Scott

In its general definition, O-CONTIGUITY militates against the insertion of an entire segment or a single feature. Given the input (14a) s-Dissimilation can apply without violating the constraint:

(16) Input = (14a)

Input: /pas-t/ O-CONTIGUITY OCP-[high] a. [past] ! b. [pa t] *!

In contrast, O-CONTIGUITY is violated given input (14b):

(17) Input = (14b):

Input: /pas-t/ O-CONTIGUITY OCP-[high] a. [past] ! b. [pa t] *!

The problem with the above alternative analysis is that it would crucially require input structures for tautomorphemic /st/ clusters which are distinct from hetero- morphemic inputs. In so doing, this analysis does not comply with Richness of the Base, which requires that multiple potential input forms be accounted for, including various autosegmental structures. More generally, the approach described above does not have a principled explanation for why (14a–b) would be required for tauto- and heteromorphemic structures respectively and not the reverse.

4 Underapplication as base-identity

4.1 Introduction

We argue below that the reason s-Dissimilation underapplies in SwG is that the stem of the suffixed form will remain faithful to the ‘base’—a unit to be defined below. Since our analysis refers to the cells in verbal paradigms, we consider in the present section some representative verb conjugations. In (18) we have provided conjugations for lachen ‘laugh’, gru¨ssen ‘greet’, essen ‘eat’ and wissen ‘know’ (from Frey 1975).10 The reason verbal paradigms have been chosen is

10 Frey does not give the complete paradigm for the second and third verbs in (18). We have modified slightly Frey’s phonetic transcriptions in (18) and below. The abbreviations in (18) and below are: INF = infinitive, 1 SG = first person singular, 2 SG = second person singular, 3 SG = third person singular, 1 PL = first person plural, 2 Pl = second person plural, 3 PL = third person plural, IMP SG = imperative singular, IMP PL = imperative plural and PASTPART = past participle. The 1–3 Sg and 1–3 Pl forms in all conjugations below are in the indicative present. We do not include the subjunctive in our analysis, but according to Frey (1975) /s/ will also surface as [s] before all subjunctive endings beginning with /t/, e.g. [viːst] ‘know (1 SG SUBJUNCTIVE II)’ (p. 132). The preterite suffix (-te [tE] in Standard German, e.g. lach-te [laxtE] ‘laughed’) does not exist in SwG (see Frey 1975, p. 127 and Vogt 1977, p. 122). In SwG past time is expressed exclusively with the past participle (e.g. er hat gelacht ‘he laughed’).

123 Inflectional paradigms have a base 161 that verbs (as opposed to words belonging to other lexical categories) show the underapplication of s-Dissimilation because many verb stems end in /s/ and two verbal suffixes are /t/ (third person singular and past participle). Put differently, there are no /t/-suffixes which could potentially attach to noun or adjective stems ending in /s/. The cells of the paradigms which involve underapplication have been shaded, while the boxed cells represent forms in which a stem-final /s/ has been deleted (see below).

(18) Verb conjugations in SwG (after Frey 1975, pp. 126–132): lachen gru¨ssen essen wissen INF. [lax-E][griEs-E][es-E] [vis-E] 1 SG [lax] [griEs] [es] [vOes] 2 SG [lax-∫][griE-∫] [i-∫] [vOe-∫] 3 SG [lax-t] [griEs-t][is-t] [vOes] 1 PL [lax-Et] [griEs-Et] [es-Et] [vis-Et] 2 PL [lax-Et] [griEs-Et] [es-Et] [vis-Et] 3 PL [lax-Et] [griEs-Et] [es-Et] [vis-Et] IMP SG [lax] [griEs] [es] [vees] IMP PL [lax-Et] [griEs-Et] [es-Et] [vis-Et] PAST PART [g-lax-t] [griEs-t][g-es-E][g-vis-t]

The first of the verbs in (18) has been given to illustrate the affixes corre- sponding to the various inflectional categories. The verbs gru¨ssen, essen and wissen are representative of verbs whose stems end in [s]. These three examples correspond to the traditional categories of weak verbs, strong verbs and irregular (weak) verbs respectively. Note the following differences between SwG and Standard German verbal conjugations: (a) The imperative singular: In SwG this member of the paradigm is always identical to the unaffixed stem, while this is not always true for strong verbs in the (e.g. [IS] ‘eat (IMP SG)’); (b) The first person singular: In contrast to the standard language, in SwG there is no suffix –e; (c) Indicative (present) plural: In SwG there is a single homophonous suffix for this category; (d) The past participle: The prefix[g] (= [gE] in the standard language) has a zero allomorph before stems beginning with a stop or an affricate; see Frey (1975, p. 130); and (e) Infinitive: The SwG infinitive is marked with an –e rather than the Standard German –en. With this background information in mind we can now consider the conju- gation of gru¨ssen in (18), paying particular attention to the phonetic represen- tation of the stem. Note that with the exception of the second person singular there is no stem allomorphy; hence, the stem is consistently pronounced [griEs] in each of the members of the verbal paradigm. There are two members of the gru¨ssen paradigm in which the stem-final /s/ is situated next to /t/, namely the third person singular and the past participle. These are the contexts in which s-Dissimilation underapplies. In the essen and wissen paradigms the cells which show underapplication of s-Dissimilation are the third person singular (for the

123 162 T. A. Hall, John H. G. Scott former verb) and the past participle (for the latter verb). Examples like these will be discussed in Sect. 4.2. An examination of the conjugation for gru¨ssen reveals the stem allomorph [griE] in the second person singular. The deletion of a stem-final sibilant in this context is purely phonological in the sense that it is triggered by a constraint banning adjacent sibilants. Hence, we can observe deletion of /s/ in the second person singular in the final two verbs in (18) as well. The details of the analysis accounting for the second person singular are spelled out in Sect. 6.

4.2 The base-identity model

We argue that the underapplication of s-Dissimilation in [s]-final stems like gru¨ssen, essen and wissen in (18) can most insightfully be analyzed by adopting the BASE-IDENTITY constraint in (19) from Raffelsiefen (1996) in her treatment of the derivational morphology of English.

(19) BASE-IDENT: The stem of the derived word must be identical to the base.

Other versions of the same constraint can be found in the literature, e.g. Kenstowicz (1996) and Kager (1999, p. 261). As we will see below, BASE-IDENT makes an analysis possible which accounts for the underapplication of s-Dis- similation by considering the surface representation as opposed to earlier levels in a derivation. The constraint in (19) makes crucial reference to a ‘base’. Following Kager (1999, pp. 281–282) and other authors, we assume that the base (with respect to inflectional paradigms like the ones in 18) is (a) a surface form, (b) a free morpheme (i.e. a word), as opposed to a bound morpheme, (c) compositionally related to its derived counterpart (i.e. the affixed form), meaning that the base contains a proper subset of the grammatical (i.e. morphological) features of the derived form. For SwG the base is simply the (present indicative) stem, i.e. the cell in the paradigm corresponding to the imperative singular (or first person singular), e.g. [griEs] for the gru¨ssen paradigm in (18).11 To see how BASE-IDENT works, we can consider the evaluation of the first person singular form (in 20) and the third person singular form (in 21) for gru¨ssen. In both tableaux we have specified the base as well as the input. See Kager (1999, p. 284–285), who takes a similar approach to the inflectional morphology of Arabic.

11 Descriptive grammars of German usually consider the base (i.e. Grundform) of the verbal conjugation to be the infinitive stem, e.g. Frey (1975, p. 129) for SwG and Eisenberg (1998, p. 146ff.) for Standard German. An examination of (18) reveals that the only verb in which the infinitive stem is not the same as the first person singular/imperative singular is wissen, which belongs to the irregular conjugation. Note that an analysis of the base as the infinitive stem does not conform to (b) above because this cell in the paradigm is not a free form. Since our analysis works in either analysis, we leave this discussion open for future research.

123 Inflectional paradigms have a base 163

(20) Input: / / BASE-IDENT OCP-[high] IO-IDENT-[high] Base: [ ] a. [ ] b. [ ] *! *

(21) Input: / - / BASE-IDENT OCP-[high] IO-IDENT-[high] Base: [ ] a. [ - ] * b. [ - ] *! *

In (20–21) we can observe that the high-ranking constraint BASE-IDENT ensures that the respective (a) candidate is optimal. The crucial example is the one in tableau (21), which shows why underapplication is the correct output: Under- application (in 21a) wins out over the ‘normal application’ of s-Dissimilation (in 21b) because of the pressure to ensure identity of the suffixed form with the base [griEs]. A few comments are in order concerning the inputs in (20–21). In contrast to the base, the input is not a surface form. In conformance with Richness of the Base, the analysis does not depend on an input with /s/; hence, the correct outputs in (20–21) will also be selected given an input with /∫/ as opposed to /S/. See also Kager (1999, p. 413ff.) and references cited therein, who acknowledges that in a model with a base the input does not play a decisive role. The base-identity approach to underapplication also accounts for verbs which have irregular stem allomorphs, e.g. the strong verb essen and the irregular verb wissen in (18). A re-examination of the wissen paradigm in (18) reveals that the base is the irregular form [vOes], but in the past participle [g-vis-t] we see a stem with a different vowel. (The third person singular of essen provides a similar example). In examples like these BASE-IDENT will select the output which incurs the fewest violations. For example, the correct output [g-vis-t] is selected over the incorrect output [g-vi∫-t] because the latter violates BASE-IDENT twice (for the vowel change and the change from [s] to [∫]), but the former one violates the same constraint only once.12

4.3 A normal application dialect

In our analysis of SwG in Sect. 4.2 we demonstrated that the ranking of constraints in (22a) correctly selects the underapplication paradigm. Were

12 Any analysis (including our own) needs to ensure that the past participle is [g-vis-t] and not [g-vOes-t]. In verbs with irregular stem alternations one needs to include the morphological context in the input, e.g. / vOes sg ~ vispl, past part / for wissen and /eS1 sg, 1–3 pl, imp., past part. ~ is2–3 sg/ for essen. This means that [g-vis-t] is selected over [g-vOes-t] because the latter form does not satisfy a con- straint which ensures that the input morphological features surface in the correct morphological context.

123 164 T. A. Hall, John H. G. Scott

BASE-IDENT to occupy a lower slot in the constraint hierarchy, then the ranking in (22b) would obtain.

(22) Two possible permutations of the constraints: a. underapplication: BASE-IDENT » OCP-[high] » IO-IDENT-[high] b. normal application: OCP-[high] » IO-IDENT-[high], BASE-IDENT As we noted above, the ranking in (22a) illustrates the unmarked pattern in the literature on southern German dialects, characterized by SwG. Thus, scores of authors have described various regional dialects spoken in southwest Germany in which under-application in verbal paradigms is the correct outcome. Given the ranking in (22b), s-Dissimilation would apply not only within morphemes but also across the stem and inflectional suffix. We refer to this ranking as ‘normal application’ because s-Dissimilation does not underapply; instead, it operates precisely where one would expect it to if it were a purely phonological process.13 We are aware of one normal application dialect of German, namely the variety of the language spoken in Eastern France (Departement Moselle) which was described by Hoffmann (1900: 29–30). We refer to this dialect below as Mosel German. In Mosel German, as in SwG and Standard German, the opposition between [s] and [∫] is neutralized to [∫] before non-high consonants in word-initial position:14

(23) Neutralization in word-initial position in Mosel German:

Mosel German Gloss schleichen [∫laixEn] ‘sneak’ schneiden [∫naidEn] ‘cut’ Schwalbe [∫molEf] ‘swallow’ ⌢ schwarz [∫toːErts] ‘black’ spalten [∫paːlEn] ‘split’ stehen [∫toːn] ‘stand’

As in SwG, the neutralization of [s] and [∫] to [∫] in Mosel German can also be observed word-internally within a morpheme before /t p/. Hoffmann does not discuss this neutralization, but it can be deduced from the examples he gives, as in (24–25). In (24) we can observe the application of s-Dissimilation within stems (pp. 37, 46, 48) and in (25) within a suffix (superlative suffix (p. 59) in (a), the ordinal suffix (p. 65) in (b) and the second person singular suffix (p. 83) in (c)):

13 In addition to underapplication and normal application, overapplication is a situation that often obtains in inflectional paradigms. A comparison of all three for the s-Dissimilation example can be found in Sect. 5. 14 Since Hoffmann uses unconventional phonetic symbols, we have replaced them with the corre- sponding characters in the IPA based on his phonetic description (pp. 4–5; 23).

123 Inflectional paradigms have a base 165

(24) Neutralization in morpheme-internal position within a stem: Mosel German Gloss Gast [ga∫t] ‘guest’ Last [la∫t] ‘burden’ Brust [broː∫t] ‘chest’ Kosten [khe∫tEn] ‘costs’

(25) Neutralization within inflectional suffixes: Mosel German Gloss a reich-ste [raix-∫t] ‘richest’ rau-ste [rau-∫t] ‘coarsest’ ⌢ b zwanzig-ste [tstansex-∫t] ‘twentieth’ c brauch-st [braux-∫t] ‘need (2 SG)’

Hoffmann (1900, p. 35) also shows that /s/ in French loan words surfaces as [∫] before /p t/, thereby demonstrating that the process was productive at the time he wrote his description. Some crucial examples are the conjugation of the verbs werfen ‘throw’, lassen ‘let’ and wissen ‘know’ in (26). The first of these three verbs illustrates the regular endings. The latter two conjugations are representative of verbs whose stem ends in /s/. Hoffmann does not say what the imperative singular is for these examples.

(26) Verb conjugations in Mosel German (Hoffmann 1900, pp. 78, 85): werfen lassen wissen INF [verf-En] [loːs-En] [ves-En] 1 SG [verf-En] [lo-En] [veːs] 2 SG [verf-∫t] [leː-∫t] [veː-∫t] 3 SG [verf-t] [leː∫-t] [veːs] 1 PL [verf-En] [loːs-En] [ves-En] 2 PL [verf-t] [loː∫-t][ve∫-t] 3 PL [verf-En] [loːs-En] [ves-En] IMP SG [verfn] —— —–

The important examples are shaded (i.e. the third person singular and second person plural of lassen and the second person plural of wissen), because the suffix for these forms is /t/. What these three forms illustrate is that, in contrast to SwG, s-Dissimilation applies across a morpheme boundary. We are assuming, based on Hoffmann (1900), that the ‘normal application’ of s- Dissimilation across a morpheme boundary in verb conjugations is the rule rather than the exception based on other examples. For example, we can observe that /s/ surfaces as [∫] in the third person singular of essen ‘eat’, fressen ‘eat’ (p. 80), blasen ‘blow’ (p. 81) and the past participle of mu¨ssen ‘must’ (p. 85).

123 166 T. A. Hall, John H. G. Scott

The upshot of the data from Mosel German is that this dialect shows that BASE-IDENTITY does not inhibit s-Dissimilation from applying. This point is illustrated in (27a), which is the verb lassen (2 Pl). We assume here that the base is [loːs], even though it is not provided by Hoffmann. Since BASE-IDENTITY is ranked low, the nature of the base is a moot point anyway.

(27) Input: /l -t/ OCP-[high] IO-IDENT-[high] BASE-IDENT Base: [l ] a. [l -t] *! b. [l -t] *

In (27) we can see that s-Dissimilation will apply across a morpheme boundary in (27b) because its underapplication in (27a) is blocked by OCP-[high]. Why are dialects like Mosel German so rare? We assume that the paucity of such dialects is an accidental by-product of history. Diachronically, the un- derapplication of s-Dissimilation in verbal paradigms was a consequence of the fact that /t/-initial suffixes were schwa-initial when s-Dissimilation entered most southern German dialects in the 13th century (Moser 1951, p. 221ff.; Russ 1978, pp. 86–87) and that for this reason there simply were no heteromorphemic examples in verbal paradigms to which s-Dissimilation could have potentially applied. Only after the historical syncope of schwa in examples like [ɡriEs-t] did the context for s-Dissimilation arise. Mosel German was innovative because it extended s-Dissimilation to examples like the one in (27b), while the numerous underapplication varieties of southern German (e.g. SwG) are more conserva- tive. We predict that there could in principle be more innovative dialects like Mosel German which are not known to us, or that such dialects could arise in the future.

5 The OP model

In the OP model of McCarthy (2005) a distinction is drawn between inflec- tional morphology, which is organized into paradigms, and derivational morphology, which hinges on a ‘derived form’. Specific OP constraints are postulated which guarantee identity with surface forms in an inflectional paradigm. An OP constraint for the feature [high] might be formalized in this approach as in (28):

(28) OP-IDENT-[high]: The value of [high] for a segment within a stem is identical to the value of the feature [high] of the corresponding segment in other words belonging to the same inflectional paradigm.

An important component of the OP model is that there is no base. This means that OP constraints like the one in (28) compare each surface form in a para- digm with every other surface form in the same paradigm. No member of the paradigm is more important than the others.

123 Inflectional paradigms have a base 167

In OP entire paradigms are candidates. Consider, for example, the paradigm given in (18) above for the verb gru¨ssen. This paradigm (the underapplication candidate) can be abbreviated as in (29a) below. Two other conceivable para- digms (in their abbreviated forms) for this verb are presented in (29b–c). The former is the ‘normal application’ paradigm because s-Dissimilation applies as expected before /t/ and nowhere else. The latter paradigm can be thought of as the ‘overapplication’ candidate because s-Dissimilation applies not only before /t/ but also in absolute word-final position.

(29) Three abbreviated paradigms for SwG. (29a) is correct and (29b–c) are incorrect. a. [griEs griEs-t…] underapplication  b. [griEs griE∫-t…] normal application  c. [griE∫ griE∫-t…] overapplication  As in our own analysis, an OP account would analyze s-Dissimilation with the ranking OCP-[high] » IO-IDENT-[high]). One might assume that the underap- plication of s-Dissimilation would require that OCP-[high] be dominated by OP-IDENT-[high], but as illustrated in (30) this ranking incorrectly predicts that the overapplication candidate should win (indicated by the backwards arrow).

(30) / ~ - …/ OP-IDENT-[high] OCP-[high] IO-IDENT-[high] a. [ ~ - …] *! b. [ ~ - …] *! * c. [ ~ - …] **

The problem with the analysis in (30) is that the overapplication candidate in (30c) is incorrectly selected as optimal because the intended winner in (30a) incurs a fatal violation of OCP-[high].15 McCarthy not only stresses that the ‘overapplication only’ prediction is correct for the Semitic examples he discusses; he also maintains that this should be the correct state of affairs for other languages as well. The reason his model makes this prediction is that both the underapplication candidate and the overapplication candidate satisfy any OP constraint, but only the latter one will satisfy the markedness constraint. McCarthy argues that underapplication in inflectional paradigms is only possible if one can motivate some high ranking (markedness) constraint which has the function of penalizing the overapplica- tion candidate so that the underapplication paradigm can be selected. His example involves underapplication in the verbal conjugation of Tiberian Hebrew (pp. 207–208): Both the underapplication and overapplication para- digms satisfy an OP constraint, but the former one wins because the latter one

15 We have yet to encounter a dialect of German in which s-Dissimilation overapplies. For this reason we do not present a formal analysis of what underapplication might look like in our own analysis.

123 168 T. A. Hall, John H. G. Scott violates a (presumably undominated) markedness constraint which prohibits schwa in the VC__CV context. To save the analysis in (30) one would therefore require a constraint which consistently penalizes (at least) one of the members of the overapplication paradigm but a fewer number of the members in the underapplication para- digm. In contrast to the Tiberian Hebrew example referred to above, it is not at all obvious what that constraint would be for the German example in (30). One option is to create a markedness constraint which penalizes an output form with [∫] in stem-final position, but this would be a classic example of an ad hoc constraint because it has no independent motivation. An added insurmountable difficulty with the markedness constraint *∫]stem is that German has many words with [∫] in the stem-final context, e.g. Rausch from (3b). Another way of saving (30) would be to appeal to a specific version of the general faithfulness constraint ANCHORING-IO. In its general form (McCarthy and Prince 1995; Kager 1999, p. 137) that constraint penalizes deletion or epenthesis at the edge of a grammatical word, but if the constraint were modified so that it also penalized a featural change (e.g. from /s/ to [∫]) in stem-final position (see van Oostendorp 2000), then (30a) would be more harmonic than (30c) because the latter paradigm would violate ANCHORING-IO twice and the former one would incur no violations. As pointed out by an anonymous reviewer, there are several reasons to believe that the ANCHORING-IO analysis is not on the right track. First, as argued by Nelson (2003), processes of reduplication require either left anchoring or the anchoring of both edges, but not right anchoring. This sug- gests that the general constraint ANCHORING-IO (right) described above might not exist as an independent constraint. Second, there is excellent evidence against the specific featural constraint ANCHORING-IO (right): In German (including the southern German varieties discussed here) left stem edges resist phonological processes much more than right stem edges. Well-known exam- ples of phonological processes affecting only a right stem edge consonant in- clude Final Devoicing and Dorsal Fricative Assimilation (i.e. the rule which accounts for the alternation between [c¸ ] and [x]). In fact, there are probably other processes in German involving the feature [high] which apply at the right stem edge (e.g. the change from /n/ to [N] before /k g/). Our conclusion is that the OP model is too strong because it does not allow for a base in inflectional paradigms. See also Albright (2004), who advocates a position similar to ours in his analysis of the historical loss of Final Devoicing in and Reiss’s (2003, p. 157ff.) criticisms of the base-identity analysis for Korean proposed by Kenstowicz (1996).

6 Explaining the second person singular

6.1 Introduction

Let us now consider the second person singular form for verbs whose stem ends in a sibilant. In (31) we have repeated the conjugation for gru¨ssen (from 18),

123 Inflectional paradigms have a base 169 which is representative of verb stems ending in /s/. We have also included here the conjugations for zischen ‘fizz’, kratzen ‘scratch’ and quatschen ‘gab’, which are representative of verbs whose stem ends in [∫] and the two coronal affricates ⌢ ⌢ ⌢ ⌢ /ts/ and /t∫/ respectively. /s∫ ts t∫ / are the only sibilants in SwG.16

(31) Verb conjugations with sibilant-final stems in SwG:

gru¨ ssen zischen kratzen quatschen ⌢ ⌢ INF.[griEs-E] [tsi∫-E] [krats-E] [kvat∫-E] ⌢ ⌢ ⌢ 1 SG [griEs] [tsi∫] [krats] [kvat∫] ⌢ ⌢ 2 SG [griE-∫][tsi-∫] [krats-E∫] [kvat∫-E∫] ⌢ ⌢ ⌢ 3 SG [griEs-t] [tsi∫-t] [krats-t] [kvat∫-t] ⌢ ⌢ ⌢ 1 PL [griEs-Et] [tsi∫-Et] [krats-Et] [kvat∫-Et] ⌢ ⌢ 2 PL [griEs-Et] [tsi∫-Et] [krats-Et] [kvat∫-Et] ⌢ ⌢ 3 PL [griEs-Et] [tsi∫-Et] [krats-Et] [kvat∫-Et] ⌢ ⌢ ⌢ IMPSG [griEs] [tsi∫] [krats] [kvat∫] ⌢ ⌢ ⌢ PASTPART [griEs-t] [tsi∫-t] [krats-t] [kvat∫-t]

A comparison of the second person singular in the four conjugations in (31) reveals that a sequence of adjacent sibilants is avoided by deleting the first one (in stems ending in /s/) or by epenthesizing schwa between the two sibilants (if the stem ends in a coronal affricate). For stems ending in [∫] it is impossible to tell whether or not the final segment of the stem or the sibilant of the suffix deletes. We are assuming the former so that we can analyze all sibilant-fricative final stems in a parallel fashion.17 An anonymous reviewer points out that we are simply assuming that the second person singular of verbs whose stem ends in a sibilant fricative (e.g. gru¨ssen) involve a deletion of the final segment of the stem (i.e. [griE-∫] from /griEs-∫/) as opposed to an assimilation of /s/ to [∫] before [∫], followed by the deletion of the suffix (i.e. [griE∫-] from /griEs-∫/). Let us refer to these two treatments as Ôanalysis AÕ and Ôanalysis BÕ respectively. We know of no empirical evidence for either approach, but it needs to be stressed that

16 The second conjugation in (31) is from Frey (1975, p. 131). Virtually none of the works cited earlier on German dialects give conjugations for verbs whose stems end in an affricate. The final two conjugations in (31) were therefore obtained from an informant. 17 See Frey (1975, p. 130), who writes that verb stems which end in /s ∫/ delete this sound before the second person singular suffix /∫/. ⌢ /, the only noncoronal affri- Neither deletion nor epenthesis occurs⌢ after a verb stem ending in /pf cate, e.g. klopfen ‘knock (2 sg)’: [klOpf∫]. This means that the pattern of deletion and epenthesis in (31) is triggered by the avoidance of adjacent (coronal) strident segments (i.e. *[CORONAL, +strident] [CORONAL, +strident] in terms of features). We do not discuss the equivalent paradigms in Standard German because s-Dissimilation never underapplies in that dialect in inflectional paradigms. (Recall from (4a–b) that s-Dissimilation only applies in Standard German word-initially). We would like to point out here that (in contrast to SwG) in the standard language the /s/ of the second person singular suffix /st/ deletes if the verb ⌢ ⌢ ⌢ stem ends in /s/, e.g. [gry:st] (from /gry:s-st/) or the affricate /ts /, e.g. kratzst [kRatst] (from / kRats-st/), while deletion does not occur when the stem ends in /∫/, e.g. [vYn∫st] (from /vYn∫-st/). We leave open what a formal analysis of this dialect would look like.

123 170 T. A. Hall, John H. G. Scott analysis B must overcome a problem not present in analysis A (which we adopt). In analysis B we have the input /griEs-∫/ mapping onto the output [griE∫-]. To account for the assimilation referred to above an additional markedness constraint would be necessary (e.g. *[s∫]) which would be ranked ahead of IDENT-[high]. The challenge is to account for why [griE∫-] is selected over the faithful form [griEs-]. This is a difficult problem to solve given the traditional (i.e. monostratal) approach to OT we are assuming here, which has an input and an output but no intermediary levels of representation. The reason the problem described here poses a challenge for the traditional OT model is that the mapping of [griE∫-] from /[griEs-∫/ involves opacity: The trigger for assimilation (i.e. the suffix /∫/) is not present on the surface. While one might be able to save analysis B by introducing a stratal-OT treatment, we do not believe that these strata derive independent support. Since analysis A does not involve opacity, it is the treatment we adopt. The processes of deletion and epenthesis in the second person singular of the verbs (31) are triggered by the markedness constraint in (32a), which penalizes a sequence of adjacent sibilants. The pattern of deletion and epenthesis follows from the interaction of *SIBSIB with the faithfulness constraints in (32b–d). Of these faithfulness constraints the first two penalize the epenthesis and deletion of any segment respectively. By contrast, constraint (32d) is a specific faithfulness constraint which penalizes the deletion of an affricate, which we have generalized here to all [–continuant] segments. The ranking required for the pattern of deletion and epenthesis in the second person singular forms in (31) is given in (32e).

(32) Constraints and rankings for 2 SG forms in SwG: a. *SIBSIB: A sequence of two sibilants is ungrammatical b. DEP-IO: No epenthesis c. MAX-IO: No deletion d. MAX-IO-[–cont]: No deletion of [–continuant] segments e. *SIBSIB,MAX-IO-[–cont] » DEP-IO » MAX-IO

The ranking in (32e) (to be illustrated in tableaux below) says that a sequence of adjacent sibilants is repaired by epenthesis rather than deletion if the first of the two sibilants is an affricate (i.e. *SIBSIB,MAX-IO-[–cont] » DEP-IO). Given a sequence of two sibilant fricatives the correct repair is to delete one of the fricatives (i.e. *SIBSIB »DEP-IO » MAX-IO). As will be shown below, there are two dialects which we define structurally in terms of the realization of the second person singular. In the first, represented by (31), the second person singular is pronounced [∫] (Sect. 6.2), and in the second (Sect. 6.3) the same morpheme is [∫t]. We refer to these two dialects henceforth as Ô[∫]-dialectsÕ and Ô[∫t]-dialectsÕ.

6.2 [∫]-dialects

The ranking in (32e) for [∫]-dialects is illustrated with the tableaux for gru¨ssen (2SG) and kratzen (2SG) in (33–34):

123 Inflectional paradigms have a base 171

(33) Input: / -/ *SIBSIB MAX-IO- DEP-IO BASE- MAX-IO Base: [ ] [–cont] IDENT a. [ -] *! b. [ -] * * c. [ - ] *!

(34) Input: / -/ *SIBSIB MAX-IO- DEP-IO BASE- MAX-IO Base: [ ] [–cont] IDENT a. [ -] *! b. [ -] *! * * c. [ - ] *

In the first tableau the deletion candidate in (33b) is correctly selected be- cause its two competitors in (33a) and (33b) violate *SIBSIB and DEP-IO respectively. This example therefore provides evidence for the ranking of the latter two constraints ahead of BASE-IDENT, which only the winner violates. By comparison, the selection of the epenthesis candidate in (34c) follows only if DEP-IO (which the winner violates) is outranked by MAX-IO-[–cont] and *SIBSIB. Although the analysis proposed in (33–34) successfully captures the facts of the second person singular for sibilant-final verb stems, it cannot explain why a stem-final consonant is deleted in (33b) and not the sibilant of the suffix. Put differently, why is [griE-∫] (2SG) in (33b) more harmonic than the incorrect [griEs-] (2SG)? Note that the incorrect candidate [griEs-] (2SG) is the one might expect to be optimal given the analysis presented in the preceding section because—in contrast to the actual winner ([griE-∫])—it does not cause the otherwise uniform paradigm to be disrupted. In tableau (34) one similarly needs to explain why the winner in (34c) is more harmonic than a candidate in which the suffix is deleted, i.e. ⌢ [krats] (2SG). It needs to be stressed that the problem posed with the pattern of epenthesis and deletion in (31) is not a theory-internal one. Traditional generative ap- proaches would require a rule deleting the first of two adjacent sibilants, but even in this analysis it is necessary to explain why it is the first sibilant and not the second one which deletes. Consider once again the gru¨ssen conjugation in (31). There are two con- ceivable explanations for why a nonoccurring form like [griEs-] (2SG) is less harmonic than the correct form [griE-∫]: (a) [griE∫s-] (2SG) is homophonous with another member of the paradigm, i.e. [griE∫s-] (IMP SG); or (b) [griE∫s-] (2SG) is an output form in which the suffix is not realized. We argue below that explanation (a) holds for [∫t]-dialects, while (b) holds for [∫]-dialects. The constraint necessary to capture explanation (b) (from Kurisu 2001, p. 39) is presented in (35).

123 172 T. A. Hall, John H. G. Scott

(35) REALIZE MORPHEME: Let a be a morphological form, b be a morphosyntactic category, and F(a) be the phonological form from which F(a+b) is derived to express a morphosyntactic category b. Then REALIZE MORPHEME is satisfied with respect to b iff F(a+b) F(a) phonologically. 6¼

The intuition behind REALIZE MORPHEME is that every morpheme receives some phonological exponence. Hence, if a suffix is deleted entirely (which would be the case in [∫]-dialects), then the constraint is violated. REALIZE MORPHEME can be illustrated by considering the conjugation of gru¨ssen ÔgreetÕ, in (31). An examination of the second person singular in this conjugation reveals that the stem-final /s/ deletes. Were the second person singular suffix to delete rather than the stem-final sibilant, then REALIZE MORPHEME would be violated. The intuition described above is captured formally in the tableau in (36):

(36) Input: / -/ REALIZE *SIBSIB DEP-IO BASE- MAX-IO Base : [ ] MORPHEME IDENT a. [ -] *! b. [ -] * * c. [ -] *! * d. [ - ] *!

Candidate (36c) is less harmonic than candidate (36b) because the former one violates REALIZE MORPHEME. Although the ranking of REALIZE MORPHEME over DEP-IO is not necessary in tableau (36), it can be deduced from verbs whose stems end in a sibilant affricate, e.g. kratzen (recall 31): The correct second person ⌢ singular form [krats- E∫] (which violates DEP-IO) is selected over the incorrect one ⌢ [kratsE-], which violates REALIZE MORPHEME.

6.3 [∫t]-dialects

The analysis presented in Sect. 6.2 will only work in [∫]-dialects. A typical example of a [∫t]-dialect is illustrated with the following conjugation of the (strong) verb essen ‘eat’ from the dialect spoken in and around Falkenberg (Tarral 1903, pp. 111–112). The verb werden ‘become’ (Tarral 1903, p. 107) has been provided as a representative example of a verb whose stem does not end in [s].18

18 In his section on strong verbs Tarral only gives the singular (indicative) forms in the conjugation in (37), so we have omitted the plural forms. Tarral does not give any examples of weak verbs ending in /s/ like gru¨ssen in (31), but the analysis we present below will go through with strong verbs as in (37).

123 Inflectional paradigms have a base 173

(37) Conjugations for werden and essen in Falkenberg German (Tarral 1903): werden essen INF. [weːr-En] [es-En] 1 SG [weːr-En] [es-En] 2 SG [weːr-∫t] [e-∫t] 3 SG [weːr-t] [es-t] IMP SG [weːr] [es] PAST PART [woːr-t] [gEsas]

In the conjugation of essen we can see that s-Dissimilation underapplies in the third person singular and that the stem-final /s/ is deleted before the second person singular suffix. The reason conjugations like the one in (37) are important is that the deletion of the stem-final sibilant in the second person singular cannot be accounted for by appealing to REALIZE MORPHEME. That constraint will not do the job because the candidate for the second person singular in which the sibilant of the suffix has been deleted (i.e. [es-t]) satisfies the constraint. We argue that the second person singular forms in dialects like the one in (37) require the anti-homophony constraint in (38) (modified slightly from Crosswhite 1999, p. 8). For a comparison of other versions of the same con- straint see Ichimura (2006) and the works cited therein. In the formalization in (38) the symbol ‘ℜ’ stands for a correspondence relation between two output strings ([S1] and [S2]). This relation produces pairs consisting of one [S1] element and one [S2] element. IO-IDENT and OO-IDENT constraints (as well as ANTI- IDENT) compare the members of these pairs and determine whether or not they are the same.

(38) ANTI-IDENT: For two forms, [S ] and [S ], where /S / /S /, $ a, a [S ], such 1 2 1 6¼ 2 2 1 that a ℜ(a). 6¼

This constraint says that for two output forms [S1] and [S2], there must be some segment a which is a member of [S1], such that a is not identical to its correspondent in [S2]. The two forms referred to here (i.e. [S1] and [S2]) are intended to be members of the same paradigm). Put simply, the constraint ANTI-IDENT penalizes an output form which has the same phonological reali- zation as another member of a paradigm. A concrete example will be discussed below. Note the requirement that the input to [S1] (i.e. /S1/) not be the same as the input to [S2] (i.e. /S2/). This condition is intended to ensure that ANTI-IDENT will allow homophony within a paradigm only if the two forms have the same input. For example, the third person singular and past participle of gru¨ssen are both [griEs-t] (recall 31), but this homophony is permitted because the two forms have the same input, namely /griEs-t/.

123 174 T. A. Hall, John H. G. Scott

The role ANTI-IDENT plays in [∫t]-dialects is illustrated in (39) for the second person singular of the verb essen in (37):19

(39) Input: / - / *SIBSIB ANTI-IDENT DEP-IO BASE- MAX-IO Base: [ ] IDENT a. [ - ] *! b. [ - ] * * c. [ - ] *! * d. [ ] *!

In this tableau we see that the deletion of the final segment of the stem in (39b) is more harmonic than the deletion of the sibilant of the suffix in (39c) because the latter candidate violates ANTI-IDENT by virtue of the fact that it is identical to the third person singular form [es-t]. To see why (39c) violates ANTI-IDENT, consider [es-t] (2 SG), which is [S1] and [es-t] (3 SG), which is [S2]. The violation of ANTI-IDENT occurs in (39c) because there is no segment a which one finds in [S1] which is not present in [S2]. Since [S1] and [S2] are segmentally identical, ANTI-IDENT is violated.20,21 Recall from note 18 that Tarral does not give the 1–3 plural cells of the essen paradigm in (37). Assuming that the second person singular is [es-t], as in Standard German, then this word would not be blocked from surfacing by ANTI-IDENT because of the /S / /S / condition in (38): 1 6¼ 2 (40) Input: / - / *SIBSIB ANTI-IDENT DEP-IO BASE- MAX-IO Base: [ ] IDENT a. [ - ] *! b. [ - ] * * c. [ - ] *! * d. [ ] *!

In this tableau the winner in (40b) satisfies ANTI-IDENT because its input is /es-t/, which is the same input for the third person singular. Put differently, the /S / /S / 1 6¼ 2

19 The tableau in (39) does not crucially require ANTI-IDENT to dominate DEP-IO. This ranking would be required assuming that epenthesis applies in the second person singular after sibilant- affricate stems (recall 31). Tarral does not give examples of verbs whose stems end in a sibilant affricate to confirm this ranking. 20 One might be tempted to explain the ungrammaticality of forms like [es-t] in (39) by replacing ANTI-IDENT with OCP-[high]. This alternative will not work because of the logic of the constraint hierarchy: Earlier in (21) it was demonstrated that underapplication is the correct outcome if BASE- IDENT outranks OCP-[high]. Since BASE-IDENT is crucially dominated by DEP-IO in (39) the impli- cation is that OCP-[high] will also be outranked by the latter constraint. 21 A superficial re-examination of tableau (36) for the weak verb gru¨ssen suggests that REALIZE- MORPHEME could be replaced with ANTI-IDENT because candidate (36c) violates both constraints. The reason this is not the correct analysis for [∫]-dialects is that it cannot account for the second person singular of strong verbs, e.g. essen `eat' (i.e. [i-∫] from /is-∫/) from (18). The second person singular form [i-∫] wins out over [is-] because [is-] violates REALIZE-MORPHEME, but crucially [is-] satisifies ANTI-IDENT.

123 Inflectional paradigms have a base 175 condition in (39) ensures that only derived homophony is avoided but that nonde- rived (i.e. underlying) homophony is allowed.

7 Conclusions and directions for further research

In this article we analyzed the canonical pattern of s-Dissimilation in SwG, in which the process underapplies before inflectional suffixes. We argued that the German facts make sense given an approach to PU which includes a base. By contrast, the data were shown to be problematic for the OP model, because that approach does not recognize a base in the evaluation of inflectional paradigms. Finally, we showed that the facts from what we call [∫]-dialects and [∫t]-dialects require constraints militating against the deletion of an entire morpheme and outputs homophonous to another member of the same paradigm (i.e. REALIZE MORPHEME and ANTI-IDENT respectively). Our treatment leaves open two general questions for further research. First, an analysis of base-identity in any language needs to say why these effects would choose to make paradigms uniform along certain dimensions (i.e. the right edge consonant in verbal conjugations) but not along others. One very obvious example of a segmental phenomenon that is typically not subject to base-identity involves the umlauted vowels in the German nominal and verbal systems, e.g. in the singular, indicative conjugation of the verb tragen ‘carry’ we have trag-e (1sg), tra¨g-st (2 sg), tra¨g-t (3 sg), trag-en (1 pl), trag-t (2 pl), trag-en (3 pl), but this is an ‘ununiform’ paradigm due to the allomorphs trag- and tra¨g-. From a formal point of view one could easily write a specific BASE-IDENT constraint for a vocalic feature (e.g. BASE-IDENT-[back]) which would be situated lower down in the constraint hierarchy, but this approach begs the question of why certain BASE-IDENT constraints are high ranked whereas other are not. We believe that an answer to this type of question needs to consider the morphological function of features like [back]. Many linguists (e.g. Wurzel 1984) have argued that umlaut in examples like the one mentioned above has a morphological (as opposed to a phonological) function. For example, in the form tra¨g-t, the morphological cat- egories Ôthird person, singular, indicativeÕ are expressed not only with the suffix –t, but also with the change in backness of the stem vowel. By contrast, the feature [high] has a purely phonological function in German consonants. What this dis- cussion therefore suggests is that the only kinds of features that cannot be present in a high-ranking BASE-IDENT constraint are those features that have some kind of mor- phological function. By contrast, features that have a purely phonological function are the only ones which can potentially be present in a high-ranking BASE-IDENT constraint. Whether or not this idea can be shown to be correct for other dialects of German or for other languages is a question we leave open for further study. A second research question is whether or not additional examples of un- derapplication in inflectional paradigms can be found. According to the OP model, underapplication should be the ‘marked’ case and overapplication the norm. We do not believe this to be the case because examples of underappli- cation abound in the . Consider, for example, the sound

123 176 T. A. Hall, John H. G. Scott change whereby the dorsal fricative /x/ dissimilated to the stop [k] before the fricative [s], e.g. vu[xs] > Modern German Fu[ks] ‘fox’, Middle High German se[xs] > Modern German se[ks] ‘six’ (Russ 1978, p. 83). Significantly, this change did not affect a stem-final [x] followed by inflectional suffixes, e.g. Modern German lach-st [laxst] ‘laugh (2 SG)’. A second example involves the rule of Compensatory Lengthening (VNC]r >VːC]r) in (Holthausen 1921), e.g. Old Saxon fu¯s ÔreadyÕ (cf. the retention of [n] in the cognate funs). Holthausen (pp. 68, 166) points out that Compensatory Lengthening is regularly blocked from applying to an [n] if an (inflectional) suffix follows, e.g. kan-st Ôcan (2 SG)Õ because the same segment is present in a morpho- logically-related word, e.g. kan Ôcan (1 SG)Õ. One could argue that these additional examples of underapplication involve only historical rules and that they therefore simply require a restructuring of the input from one stage to the next, rather than a BASE-IDENT constraint (Reiss 2003). Another possibility is that—at least for some dialects—these historical rules represented synchronic rules, in which case an analysis with BASE-IDENT might do more justice to the facts.

Acknowledgements

We would like to thank the audiences at the Phonology Fest at Indiana University in June 2006, the Zentrum fu¨ r Allgemeine Sprachwissenschaft (ZAS) in Berlin in February 2007 and the Fifteenth Manchester Phonology Meeting in May 2007 for helpful comments on earlier versions of the present article. In particular, we would like to thank Laura Downing, Barbara Stiebels, Jochen Trommer, Marzena Zygis for their helpful questions. We would also like to thank two anonymous referees for their thoughtful feedback. Any remaining errors are our own responsibility.

References

Alber, B. (2001). Regional variation and edges: Glottal stop epenthesis and dissimilation in standard and Southern varieties of German. Zeitschrift fu¨r Sprachwissenschaft, 20, 3–41. Albright, A. (2004). Sub-optimal paradigms in Yiddish. WCCFL, 23, 1–14. Benua, L. (1997). Transderivational identity: Phonological relations between words. Ph.D. disser- tation, Amherst: University of Massachusetts. Berger, J. (1913). Die Laute der Mundarten des St. Galler Reintals und der angrenzenden vora- rlbergischen Gebiete. Betra¨ge zur Schweizerdeutschen Grammatik, 3, 1–231. Bertram, O. (1937). Die Mundart der mittleren Vorderpfalz. Erlangen: Palm & Enke. Bohnenberger, K. (1928). Die Mundarten Wu¨rtenbergs. Eine heimatkundliche Sprachlehre. Stuttgart: Silberburg. Booij, G. (1985). Coordination reduction in complex words: a case for prosodic phonology. In H. van der Hulst & N. Smith (Eds.), Advances in nonlinear phonology (pp. 143–160). Dordrecht: Foris. Clements, G. N. (1999). Affricates as noncontoured stops. In O. Fujimura, et al. (Eds.), Item, order in Language and Speech (pp. 271–299). Prague: Charles University Press. Crosswhite, K. (1999). Intra-paradigmatic homophony avoidance in two dialects of Slavic. In Matthew K. Gordon (Ed.), UCLA working papers in linguistics 1. Papers in phonology 2. UCLA. Downing, L., Hall, T. A., & Raffelsiefen, R. (Eds.). (2005). Paradigms in phonological theory. Oxford: Oxford University Press.

123 Inflectional paradigms have a base 177

Eisenberg, P. (1998). Grundriss der deutschen Grammatik. Stuttgart: J. B. Metzler. Fleischer, W., & Barz, I. (1995). Wortbildung der deutschen Gegenwartssprache (2nd ed.). Tu¨ bin- gen: Niemeyer. Frey, E. (1975). Stuttgarter Schwa¨bisch. Laut- und Formenlehre eines Stuttgarter Idiolekts. Mar- burg: N. G. Elwert. Grijzenhout, J. (1998). The role of coronal underspecification in German and Dutch phonology and morphology. In W. Kehrein & R. Wiese (Eds.), Phonology and morphology of the (pp. 27–50). Tu¨ bingen: Max Niemeyer. Haag, C. (1898). Die Mundarten des oberen Neckar- und Donaulandes. Reutlingen: Eugen Hutzler. Hall, T. A. (1992). Syllable structure and syllable related processes in German. Tu¨ bingen: Niemeyer. Hall, T. A. (2002). The distribution of superheavy syllables in Standard German. The Linguistic Review, 19, 377–420. Hall, T. A. (2007). Middle high German [rs] > [r∫] as height dissimilation. Ms. Indiana University. Hoffmann, K. (1900). Laut- und Flexionslehre der Mundart der Moselgegend von Oberham bis zur Reinprovinz. Metz: Buchdruckerei der Lothringer Zeitung. Holthausen, F. (1921). Alsa¨chsisches Elementarbuch. Heidelberg: Carl Winter’s Universita¨ ts- buchhandlung. Ichimura, L. (2006). Anti-homophony blocking and its productivity in transparadigmatic relations. Ph.D. dissertation: Boston University [ROA–881–1006]. Itoˆ , J. & Mester, A. (1999). Realignment. In R. Kager, H. van der Hulst, & W. Zonneveld (Eds.), The prosody-morphology interface. (pp. 188–217). Cambridge: Cambridge University Press. Jakobson, R., Fant, G., & Halle, M. (1952). Preliminaries to speech analysis: The distinctive fea- tures and their correlates. Cambridge: MIT Press. Jutz, L. (1931). Die alemannischen Mundarten (Abriss der Lautverha¨ltnisse). Halle (Saale): Max Niemeyer. Kager, R. (1999). Optimality theory. Cambridge: Cambridge University Press. Kehrein, W. (2002). Phonological representation and phonetic phrasing: Affricates and laryngeals. Tu¨ bingen: Niemeyer. Kenstowicz, M. (1996). Base-identity and uniform exponence: Alternatives to cyclicity. In J. Durand & B. Laks (Eds.), Current trends in phonology: models and methods (pp. 363–393). Man- chester: European Studies Research Institute. Kenstowicz, M. (2005). Paradigmatic uniformity and contrast. In L. Downing, T.A. Hall, & R. Raffelsiefen (Eds.), Paradigms in Optimality Theory (pp. 145–169). Oxford: Oxford University Press. Kim, H. (2001). A phonetically based account of phonological stop assibilation. Phonology, 18, 81– 108. Kurisu, K. (2001). The phonology of morpheme realization. Ph.D. dissertation, University of California at Santa Cruz [ROA-490–0102]. LaCharite´ , D. (1993). The internal structure of affricates. Ph.D. dissertation, University of Ottawa. Lahiri, A., & Evers, V. (1991). Palatalization and coronality. In C. Paradis & J.-F. Prunet (Eds.), The special status of coronals. Internal and external evidence (pp. 79–100). San Diego: Aca- demic Press. McCarthy, J. (2005). Optimal paradigms. In L. Downing, T. A. Hall, & R. Raffelsiefen (Eds.), Paradigms in phonological theory (pp. 170–210). Oxford: Oxford University Press. McCarthy, J., & Prince, A. (1995). Faithfulness and reduplicative identity. In J. Beckman, L. Walsh Dickey, & S. Urbanczyk (Eds.), Papers in optimality theory (pp. 249–384). University of Massachusetts Occasional Papers in Linguistics 18. Amherst, Mass. Moser, V. (1951). Fru¨hneuhochdeutsche Grammatik. Heidelberg: Carl Winter Universita¨ tsverlag. Muthmann, G. (1988). Phonologisches Wo¨rterbuch der deutschen Sprache. Tu¨ bingen: Niemeyer. Nelson, N. A. (2003). Asymmetric anchoring. Ph.D. dissertation, Rutgers University. Nı´ Chiosa´ in, M. (1991). Topics in the phonology of Irish. Ph.D. dissertation, Amherst: University of Massachusetts. Noske, M. (1997). Feature spreading as dealignment: the distribution of [c¸ ] and [x] in German. Phonology, 14, 221–234. Prince, A., & Smolensky, P. (2004). Optimality theory: Constraint interaction in generative gram- mar. Malden, Massachusetts: Blackwell. Raffelsiefen, R. (1995). Conditions for stability: The case of schwa in German. Arbeiten des Son- derforschungsbereichs 282. Du¨ sseldorf: Heinrich-Heine-Universita¨ t.

123 178 T. A. Hall, John H. G. Scott

Raffelsiefen, R. (1996). Gaps in word formation. In U. Kleinhenz (Ed.), Interfaces in Phonology (pp. 194–209). Berlin: Akademie Verlag. Raffelsiefen, R. (2004). Phonological Effects in Word formation. Habilitationsschrift. Freie University of Berlin. Raffelsiefen, R. (2000). Evidence for word-internal phonological words in German. In R. Thieroff, et al. (Eds.), Deutsche Grammatik in Theorie und Praxis (pp. 43–56). Tu¨ bingen: Niemeyer. Reiss, C. (2003). Language change without constraint reranking. In D. E. Holt (Ed.), Optimality theory and language change (pp. 143–168). Dordrecht: Kluwer. Rubach, J. (1994). Affricates as strident stops in Polish. Linguistic Inquiry, 25, 119–143. Russ, C. V. J. (1978). Historical German phonology and morphology. Oxford: Clarendon Press. Russ, C. V. (1982). Studies in historical German phonology. Bern: Peter Lang. Schatz, J. (1897). Die Mundart von Imst. Laut- und Flexionslehre. Strassburg: Karl J. Tru¨ bner. Schirmunski, V. M. (1962). Deutsche Mundartkunde. Vergleichende Laut- und Formenlehre der deutschen Mundarten. Berlin: Akademie-Verlag. Scott, J. H. G. (2006). Dissimilation of coronal fricatives in Swabian German tautomorphemic /sC/ clusters. Ms. Indiana University. Steriade, D. (2000). Paradigm uniformity and the phonetics-phonology boundary. In M. Broe & J. Pierrehumbert (Eds.), Papers in laboratory phonology V. Acquisition and the lexicon (pp. 313–334). Cambridge: Cambridge University Press. Tarral, N. (1903). Laut-und Formenlehre der Mundart des Kantons Falkenberg in Lothr. Strass- burg: Heitz & Mu¨ ndel. van Oostendorp, M. (2000). Wieringse nasaalvelarisiering. Taal en Tongval LLL, 1, 163–188. Vogt, E. F. (1977). Schwa¨bisch in Laut und Schrift. Stuttgart: J. F. Steinkopf. Wiese, R. (1991). Was ist extrasilbisch im Deutschen und warum? Zeitschrift fu¨r Sprachwissen- schaft, 10.1, 112–133. Wiese, R. (1996). The phonology of German. Oxford: Clarendon Press. Wipf, E. (1910). Die Mundart von Visperterminen im Wallis. Betra¨ge zur Schweizerdeutschen Grammatik, 2, 1–198. Wurzel, W. U. (1984). Flexionsmorphologie und Natu¨rlichkeit. Berlin: Akademie-Verlag. Zinser, R. (1933). Die Mundart des Oberen Ga¨us su¨dlich von Herrenberg nach Lauten und Flexion. Stuttgart: J. Fink.

123