• Abstracts

Evangelia Adamou (Lacito, CNRS)

From conceptual transfer to grammatical replication:The influence of the Spanish copulas on Romani spoken in Latin America   

When two languages in contact have different conceptual representations encoded by distinct linguistic means, conceptual transfer is likely to occur from an L1 to an L2 and from a dominant language to a heritage or minority language (Jarvis and Pavlenko 2008 for an overview). For example, it is widely documented that English speakers who have a single conceptual representation of being, expressed by the copula to be, encounter difficulties in the use of the Spanish copulas ser and estar which correspond to two conceptual representations. In particular, several studies show that English L2 learners of Spanish generalize ser before acquiring estar (e.g., Geeslin 2001), and that heritage speakers of Spanish in the US generalize estar (Silva-Corvalán 1986) following a more general trend among Spanish monolinguals in the Americas (e.g., Gutiérrez 1994 for Mexico, Ortiz-López 2000 for Puerto Rico).  

In Adamou (2013) I presented evidence from a previously undocumented contact setting involving Mexican Spanish and heritage Romani (Indic, Indo-European) that illustrate the long-term effects of partial conceptual equivalence in the absence of normative pressure. Specifically, whereas Romani speakers from Europe utilize a single copula (Matras 2002, Elšik and Matras 2006), speakers of Romani who are settled in Mexico since the 19th century distinguish between attributive predications using the copula si, as in (1a), and the third person subject clitic pronouns, as in (1b).

(1) Romani from Oaxaca, Mexico
     a.    le          ʃave       muᴚa         bibiake     si          barbale
            DEF.PL  children  POSS.1SG  aunt.DAT  be.3PL   rich
          ‘My auntʼs children are rich.’(Adamou 2013: 1085)
     b.   o         raklo=lo      felis
          DEF.M  boy=3SG.M  happy
         ‘The boy is happy.’ (Adamou 2013: 1075)

I suggest that heritage Romani speakers developed two conceptual representations of being to parallel those of Spanish and then recruited obsolescent material in Romani, i.e. the subject clitic pronouns, to replicate the uses of estar. A recent study based on quantitative data from Romani spoken in Veracruz, Mexico sheds more light on this equivalence and allows for the elaboration of a more fine-grained diachronic scenario (Padure, in progress). Finally, in addition to the Mexican data, Acuña and Adamou (2013) reported similar uses from Romani spoken in Bogota, Colombia, indicating that this innovation is a widespread Romani feature in Latin America.

Isabelle Buchstaller (U Leipzig)

Exploring transfer (including metatype) phenomena in Marshallese English

This paper is a first report from an ongoing ethnographic research project on post-WWII language contact in the Republic of the Marshall Islands (RMI). I present a sketch of the varieties that resulted from contact between English and Marshallese, the first language of the inhabitants of the Republic of the Marshall Islands. I will outline some attested characteristics of Marshallese English (ME), investigating their possible relationship to Marshallese as transfer phenomena. In particular, I will focus on the English verb phrase, which reveals ‘metatypy effects’ in the sense of Ross (2006).

Mariapaola D'Imperio (U Aix-Marseille)

Dialect imitation across typologically distinct prosodic systems

During the process of second language acquisition, both L1 phonemic and prosodic categories might be subject to transfer from the L1 to the L2. Among prosodic features, the early-acquired metrical and intonational properties of the L1 might shape L2 prosody and be partly the basis of the perception of Foreign Accent (FA). On the production side, speakers attempting to imitate an unfamiliar system must learn which factors govern the variability present in the target speech. This is expected to be more difficult when the speaker’s native system and the target system are typologically distinct (Best 1995). Nevertheless, speakers can adjust phonetic details of their own segmental pronunciation so that they can narrowly resemble other speaker’s productions, which they have just heard (Goldinger 1998, Nielsen 2007, 2010, inter alia), though studies of prosodic and intonational imitation have presented mixed results. On one side, it has been argued that speakers are only able to reproduce gross phonological patterns (Cole & Shattuck Hufnagel 2011) or else that both phonological and fine phonetic detail (such as tonal alignment and scaling) can be successfully
reproduced (D’Imperio, Petrone & Cavone 2014).

Both metrical and intonational typology can also vary among varieties of the same language, as in the case of Southern and Standard French (Selkirk 1977). At the foot level, languages such as English and Italian are trochaic, while a language such as Standard French shows an iambic structure (Hayes 1995). Also, at an intonational/typological level, while American English is a head language (Jun 2005), Singapore English is an edge language. Given that learning an unfamiliar dialect or an L2 can be thought of as a process of long-term imitation, the studies reviewed here will be based upon the direct-imitation paradigm (German, Carlson & Pierrehumbert 2013).

Hence, in this talk I’ll summarize 3 studies concerning learners’ imitations of either metrical or intonational properties of a typologically different language (French vs. Italian) or a different variety of the same language (Singapore vs. American English and Southern vs. Standard French). Given that the amount of L1 use can directly impact perceived FA, at least in segmental production in late and advanced L2 learners (Flege et al., 1997), in study 1 we tested whether L1 use would also result in greater prosodic FA both at the production and the perception level (Cavone & D’Imperio, submitted). We also know that larger exposure to speech productions can lead to stronger episodic traces (Goldinger, 1998), hence we might hypothesize that increased L2 exposure can lead to better target productions on the part of the learners, which was tested in study 2 (D’Imperio & German, 2015, submitted).

Our results show that, despite the structural difference, speaker of typologically different languages can rapidly imitate fine phonetic detail either related to metrical structure or intonational structure. Specifically, for the French L2 study, we show that L1 use is a predictor of how well the Italian learners’ imitation process will apply as well as their generalization performance. Also, speakers of Singapore English are able to rapidly adapt and shift from an edge-based system to an accentual system within the time of the experiment, as well as finely tune intonation phonetic detail of tonal alignment so as to imitate a model American speaker’s pronunciation. Finally, the degree of variability in successfully reproducing the target values appears to be also dependent on amount of exposure to the non-native, L2, dialect. I’ll then discuss the results in terms of a general model of L2 prosodic acquisition.

Hubert Haider (U Salzburg)

Do not copy & paste  - No replications in syntactic derivations

Replication-by-derivation is a core concept within the Minimalist Program. Covert replication is a  consequence of the so-called ’copy theory of movement’. In the MP, ‘movement’ is a complex device based on two crucial assumptions. The first assumption is the copy & paste assumption. In structures with displaced items, the derivational path starting at the base position and ending at the final position is deemed to be paved with pasted copies of the ‘moved’ items. The second assumption is a ’spell-out’ convention. Only one of the copies gets spelled out. In English, it is typically the structurally highest copy. The spell-out convention owes its empirical footing mainly to the cross-linguistic comparison of wh-movement-constructions (overt vs. covert movement options) but this evidence turns into counterevidence when properly investigated.

This presentation will expound why the empirical basis in support of covert-movement as well as a copy theory of movement is insufficient or inexistent, and how serious counterevidence renders both assumptions highly implausible.

Fabian Heck (U Leipzig)

Order perservation without reference to linear order — the case of Holmberg's generalization

Holmberg (1986) observes that an object may raise out of the vP in Scandinavian only if the verb does so, too. At least two general proposals may be identified as to how Holmberg's generalization (HG) should be accounted for. The first approach has it that HG goes back to a mechanism that is designed to preserve (``copy'') the relative order of verb and object (Fox & Pesetsky 2005, Broekhuis 2008, Engels & Vikner 2014). The second approach puts forward the idea that object shift poses a locality problem for subject raising, and that somehow verb movement can solve this problem (Chomsky 1993, Bobaljik & Jonas 1996, Kitahara 1997). In my presentation, I will propose a new version of the second approach, arguing that the idea that verb movement is required to enable subject raising in the context of object shift can be derived from a new type of derivation (which I call ``non-monotonic'') that mostly rests on independently motivated assumptions.

Andrew Kostakis (U Leipzig)

Spreading height features in Germanic metaphony

In Germanic (Gmc.) languages, mid vowels pattern as complex segments that are both [high] and [low]. This claim is corroborated by mid vowels that unpack their [high] and [low] features into two separate [high] and [low] segments (e.g. [ɛ] > [ja]). It is also supported by mid vowels that obtain from the coalescence of distinct [high] and [low] segments (e.g. [au] > [ō]). Metaphonic processes in Gmc. languages furnish additional evidence for complex mid vowels. The talk will focus on two such changes. The first is OHG Primary Umlaut, a sound change that raised and fronted a tonic [a] to [e] before a following [i] or [j]. Alternations in forms like OHG l[a]mb ‘lamb’ ~ l[e]mbir ‘lambs’ exemplify the change. I argue that the new OHG mid vowel resulted from [high]-spreading: the feature [high] from the umlaut trigger spread onto the tonic low vowel and produced a ([high] and [low]) mid vowel. The second change is Northwest Gmc. (NWGmc.) a-Umlaut, the lowering of a tonic [u] to [ɔ] preceding a non-high vowel. NWGmc. a-Umlaut shifted the high vowel in examples like PGmc. *mulđōn ‘mold’ to a mid vowel in ON (cf. mold) and OHG (cf. molta). The change is analyzed as [low]-spreading, whereby the feature [low] of the umlaut trigger spreads onto the preceding high, tonic vowel. Gmc. Metaphony is interesting because certain consonants block its application. OHG Primary Umlaut is blocked by [x] and the coda liquids [r]σ and [l]σ; NWGmc. a-Umlaut is blocked by a coda nasal. I argue that the no-crossing constraint is responsible for these blocking effects. Thus, OHG Primary Umlaut – a process of [high]-spreading – is blocked by a natural class of [high] consonants ([x], [r]σ and [l]σ), while NWGmc. a-Mutation – which results from [low]-spreading – occurs because coda nasals are marked with [low]. Independent evidence for the proposed consonantal feature specifications can be observed outside of Gmc. languages.

Gereon Müller (U Leipzig)

Copies without Copy Theory

In this talk, I argue for an approach to copy constructions in syntax that does not rely on the copy theory of movement; rather, it employs copy operations as in standard cases of reduplication. As a case study, I consider predicate doubling constructions in (two varieties of) German, Hebrew, and Asante Twi. I argue that the choice between copying and do-support in predicate doubling constructions can be reduced to the order of the two operations Move and Copy. The analysis will then be extended to two further copy constructions in German: wh-copying and replicative idioms.

Andrew Nevins (UC London)

Copying and resolution in South Slavic and South Bantu conjunct agreement

In Willer-Gold et al (2016), on the basis of asymmetries between Neuter and Feminine Gender in Conjunct Agreement in Slavic, a distinction is proposed between default agreement on &P (a fixed value of masculine plural, independent of the values of either conjunct), and resolved agreement, a computation that depends on the values of each conjunct and their relation to each other. On the basis of our experimental results, the following input-output functions are proposed for South Slavic:

(1) Gender Resolution outcomes:
     M&F, F&M → M
     M&N, N&M → N
     F&N, N&F → N

Specifying (1) is important, but what is the mechanism that yields these, and how does it relate to other computations in the grammar? Pušskar & Murphy (2015) speculate that resolution is achieved by impoverishment, a  view we take up here, though depending on a different feature set, in which neuter is least specified (and masculine and feminine even when inanimate and hence uninterpretable are both specified for [+common]).

By contrast, Dalrymple & Kaplan (2000) propose a set-theoretic union operation to derive resolution rules, whereby Fem is represented as an empty set ={ }, thus a resolution rule whereby {M} ∪ { } = {M}, but this representation does not square with the markedness relations in South Slavic. For three-gender languages with the resolution pattern of Icelandic, they propose that Neuter is composed of a set {M,F}, which predicts that Masc+Fem (i.e. {M} ∪ {F} = {M,F}), will yield Neuter. Unlike Icelandic, in South Slavic, M&F yields M. There is no way to achieve a two-level theory of resolution (like ours, where M&F = M but M/F & N = N) with privative features and set-theoretic union. For Slovene, they argue, based on Corbett 1983: 186, that N&N = M. Our results show little support for this, as N&N also yields N (and much more so than M). On the other hand, set-intersection (instead of union) would also founder, as N&N would yield N only, and not allow M. Recall that under our model, the latter pattern is actually not resolution (as Dalrymple & Kaplan would have it), but default. Dalrymple & Kaplan (2000) in fact suggest something like this: that, on the basis of parallel patterns with noun class resolution in Lama (Yu 1988), the coordination head itself provides an additional gender value.

The patterns of gender resolution in South Slavic are further enriched by considering Southern Bantu, on the basis of data reported in Mitchley (2015). Of particular interest is the ineffability that occurs when human and non-human nouns are conjoined, even when they are of the same class, posing particular problems for an OT approach to selecting the optimal candidate.

We propose that gender resolution takes place via a bidirectional search-and-copy mechanism with varying prespecification on the &P head, taking the vowel-harmony analysis of Nevins (2010) for Woleaian and applying it back to phonological concord. Masculine and feminine both bear [+common], while feminine bears [+fem]. Only values found on both sides can be kept. In Southern Bantu, however, both [+human] and [–human] are equipollently specified (in addition to other gender features), leading to an inevitable clash.

Number resolution takes place in apparently a wholly different manner, and one that cannot be achieved by impoverishment or union/intersection of singular features. Instead, we liken number resolutions mechanisms to those proposed by Trommer (2006) and Gluckman (2015). Due to the different nature of the resolution mechanism, we propose that number resolution must come before gender resolution, leading to the ‘consistency principle’ results whereby gender agreement depends on prior number resolution, as reported in Marušič et al (2015) and in Mitchley’s data for isiXhosa, potentially obviating the need to extrinsically order these two.

Douglas Pulleyblank (U of British Columbia)

Reduplication in Nuu-chah-nulth

In this talk I will consider nine reduplicative patterns attested in Nuu-chah-nulth, a Wakashan language of British Columbia. Each of the nine patterns is associated with a particular morphological property, sometimes occurring in conjunction with a suffix and sometime occurring as the sole exponent of the morphological property in question. All reduplicative patterns are word-initial; all involve a single syllable. Some of the patterns copy length while others are systematically long or short; some of the patterns copy codas while others invariably involve open-syllabled reduplicants. Some of the patterns involve modifications of the first vowel of the root while others do not.

The richness of the Nuu-chah-nulth patterns poses interesting problems analytically. From a templatic perspective, the number of patterns raises issues about both the number and nature of reduplicative templates. From an ‘A-templatic’ or generalized template perspective, the patterns pose problems given the need to differentiate among the various patterns, whether viewed as affixing or compounding. The proposal made in the talk is to account for Nuu-chah-nulth reduplication using a surface-oriented ‘emergent’ approach.

Hannah Sande (UC Berkeley)

Apparent weight-dependent infixing reduplication in Amharic

In this talk I describe the infixing reduplication pattern of Amharic adjectives and verbs based on both original and existing data. I demonstrate that a Stratal OT account of the data is superior to a parallel or Harmonic Serialism one in that it avoids stipulating an otherwise unattested infixation target (geminates or heavy syllables). With a stratal account we see that infixing reduplication in Amharic targets heavy syllables at the stem level.

Donca Steriade (MIT)

Grammars of rhyme: how they differ and how they are the same

An analysis of cca 60 poetic traditions shows that rhyming techniques differ substantially from one system to the next. Differences observed involve the position of the Rhyming Domain (RD) within the line; the extent of required similarity between paired RDs; and minimal size and the presence of a metrically prominent position inside the RD.

Rhyming grammars however converge on one point: the position of the left and right edges of the  RD. These are arguably invariant: the RDs begin with a nucleus and they end either before a nucleus (when the RD is word internal) or with the end of the relevant metrical domain.

To explore what causes rhyming systems to diverge in limited ways, I analyze them as OT grammars, i.e. as systems that evaluate linguistic expressions, returning a judgment about the wellformedness of each potential expression (here, each potential rhyming pair), based on a hierarchy of considerations. I argue that the invariant properties of RDs arise from the internal structure of the rhythmic units that the line of verse is built from.

Barbara Stiebels (U Leipzig)

Multiple exponence in a lexical-incremental framework

Unlike theoretical approaches to morphology that foresee the occurrence of multiple exponence (ME), e.g., Anderson (2001), Stump (2001), lexical-incremental approaches such as Minimalist Morphology (MM; Wunderlich & Fabri 1995) ban ME by a general anti-redundancy constraint, which is meant to rule out massive over-generation of morphological forms in paradigm construction. I will discuss which alleged cases of ME can be reanalyzed easily within MM (updating and extending Ortmann 1999), and which data pose a serious challenge for MM.

Sandhya Sundaresan (U Leipzig)

Replicative processes in reference and their failures: pro vs. PRO

Whenever two related grammatical elements display distinct properties and appear in distinct environments, we have in principle two analytic possibilities:

  • We can posit distinctions inherent to the elements themselves, which then lead to them having different distributions (the Inherent View)
  • We can posit a single underlying element with a broad distribution, with the grammatical distinctions arising from the different contexts in which that element appears (the Derived View)

One place where this analytic tension is played out is in the analysis of non-overt subjects in different contexts, what are traditionally called obligatorily controlled (OC) “PRO" and (little) “pro". These two elements have a great deal in common, with both being silent, (pro-)nominal, and occurring typically in subject position. Yet they are systematically distinguished with respect to their interpretive properties, with the element called “PRO” behaving fundamentally like a bound variable and that called “pro” behaving essentially like a pronoun. It is probably safe to say that it is standard to take an Inherent approach to this distinction (e.g. Chomsky 1981, Martin 2001, Hornstein 1999, Landau 2004, 2013), with PRO being treated as being featurally anaphoric and pro being featurally pronominal. Here, I will instead pursue a Derived approach, arguing as follows. There is a single referentially deficient element (“UPro”) which underlies the labels “PRO” and “pro”: the differences between the two forms are derived by interactions with the syntactic structure which lead to different syntactic processes of the replication of reference. In particular,  OC PRO involves replication from an overt locally c-commanding DP, different types of pro, and NOC PRO, involve replication from usually silent representations of discourse-contextual elements in the left periphery, and arbitrary PRO arguably involves the failure to replicate reference (formalised as the failure to Agree in the sense of Preminger (2014)). Empirical evidence for the Derived approach proposed here will be adduced from the "Finiteness pro-drop Generalisation” (Sundaresan 2014) which shows that interpretations PRO and pro are in strictly complementary distribution in a way that goes beyond what standard Case based theories of PRO would lead us to expect. A Derived approach has a way to deal with this since PRO and pro are treated as being contextual variants of the same underlying element; but an Inherent approach would have to stipulate something extra to derive this.

Jochen Trommer (U Leipzig)

Featural affixation and featural spreading in Gaahmg

In the autosegmental approach to morphophonology, the operations and constraints governing featural spreading and featural affixation are almost completely coextensive. This makes the strong prediction that both processes should work in parallel crosslinguistically, and especially in single languages.  In this talk, I show that this prediction is largely borne out in the vocalic and tonal morphophonology of Gaahmg, a Nilotic language spoken in Southern Sudan  (Stirtz 2011).  Gaahmg is especially interesting for this question because it shows a highly heterogeneous set of featural affixation processes (including pure addition of features, complete overwriting of base features by affixal features, and partial overwriting), but also because the major tonal spreading  process in the language is subtonemic, i.e. consists in spreading a lower tonal register, not entire tones. I show that both, register spreading and partial tonal overwriting follow straightforwardly from the decomposition of tonal features in Register Tone Theory  (Snider 1999) and converge in reflecting the stratal organization of morphophonology suggested by Stratal Optimality Theory  (Bermúdez-Otero in preparation, Trommer 2011).

Stephen Wechsler (U of Texas at Austin)

Explaining Grammatical Agreement

Grammatical agreement develops when the covariation of forms in a sentence due to coreference is reanalyzed as resulting from a rule of morphosyntax. Two main types of coreference are involved: pronoun-antecedent relations, from which person-rich (Index) agreement derives; and the repetition of classifiers on a noun head and its modifiers, from which person-poor, case-rich (Concord) agreement derives. This talk explores some ways grammatical properties of agreement systems can be explained in terms of their semantic provenance.  (i) Locality: For a rule to be induced by the learner, the covariation must be systematic, hence there must be a consistent grammatical dependency between the controller and the target.  (ii) Formal and semantic agreement: The semantic content of the target features is rarely lost entirely; it emerges whenever evidence of a rule is lacking because the controller is unspecified for the feature (the Agreement Marking Principle, Wechsler 2011).  (iii) Feature sharing: When the controller assumes properties of the target, specific predictions follow.  First, in headedness reversal, Concord agreement gives the appearance of long distance agreement (Haug and Nikitina 2015).  Second, when the controller of Index agreement is a pronoun it is necessarily marked for Index features, blocking semantic agreement.

Richard Wiese (U Marburg)

How language can be as regular as music

Language seems to contrast with music (at least with classical Western music) by being less regularly in terms of metrical organization. In this contribution, I argue that language, contrary to the obvious impression, follows the same preference for a regular metrical organization as music. In order to substantiate this hypothesis, I present the results of an extensive corpus study of spoken and written German. The specific phenomenon studied is that of schwa-zero alternations in German. For a large range of words, such alternations exist, as in gern - gerne ‘gladly'. These words make it possible to pursue the hypothesis of Prosodic Parallelism.

Eva Zimmermann (U Leipzig)

Multiple reduplication as an argument against the Red-morpheme

In this talk, I argue that the intricate pattern of multiple reduplication in Lushootseed follows best in an account that takes reduplication as the result of a general phonological copying operation that avoids an otherwise phonologically marked structure. Lushootseed (Broselow, 1983; Bates, 1986; Urbanczyk, 1999, 2001) employs two differently-sized reduplication patterns: whereas a prefixed /Ci/-reduplicant marks the diminutive (1b), the distributive is marked by a prefixed /CVC/-sequence (1c). Interestingly, both these reduplicants can cooccurr in both orders DIM-DIST (1d) and DIST-DIM (1e). The theoretically challenging fact is now that the distributive reduplicant in (1e) apparently only ‘sees’ the adjacent reduplicant, not the base, and can hence only copy a CV-string, not a CVC-string as in all other contexts. This has led to arguments that this pattern requires a cyclic analysis and the assumption of a subjacency condition (Broselow, 1983) or the assumption of different morphological statuses for the two reduplicating morphemes (Urbanczyk,

(1) Reduplication in Lushootseed (Broselow, 1983, 319-325)
     a. bədáʔ          ‘child’
     b. bíbədaʔ       ‘small child
     c. bədbədáʔ     ‘children’
     d. bíbədbədaʔ  ‘dolls, litter’
     e. bibibədaʔ    ‘small children’

It is shown that this pattern straightforwardly follows from the basic mechanism of affixation of nonsegmental
affixes (prosodic nodes or features) and copying of underlying material as one phonological strategy to fill those otherwise empty affixal nodes with material (=Theory of Minimal Reduplication, Saba Kirchner, 2007, 2010). The apparent ‘invisibility’ of the base in (1e) is shown to be an epiphenomenon under a phonological account for the facts that takes the two copying-triggering morphemes to be smaller than in any alternative analysis: the copied consonants in (1e) are in fact not copied to fill a reduplicant but to provide otherwise onset-less syllables with an onset-consonant. Such a general account of reduplication as non-segmental affixation can also easily predict the full typology of attested multiple reduplication patterns: multiple reduplicants in all contexts as in Lushootseed, multiple reduplication only if it has some other independent motivation as in Chaha (Rose, 1997), and avoidance of multiple reduplicants and surfacing of one ‘superset’-reduplicant in Nuuchahnulth (Stonham, 2004, 2007). This latter avoidance of multiple adjacent ‘reduplicants’ as in Nuuchahnulth is in this theory simply the coalescence of affixal prosodic nodes to avoid abundant copying of base material.

letzte Änderung: 06.01.2017